lectures/21_communication.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "![](img/330-banner.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "-"
    }
   },
   "source": [
    "# Lecture 21: Communication\n",
    "\n",
    "UBC 2022 Summer\n",
    "\n",
    "Instructor: Mehrdad Oveisi"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "## Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from sklearn.compose import ColumnTransformer, TransformedTargetRegressor\n",
    "from sklearn.dummy import DummyClassifier, DummyRegressor\n",
    "from sklearn.ensemble import RandomForestClassifier, RandomForestRegressor\n",
    "from sklearn.impute import SimpleImputer\n",
    "from sklearn.linear_model import Ridge\n",
    "from sklearn.metrics import log_loss\n",
    "from sklearn.model_selection import (\n",
    "    GridSearchCV,\n",
    "    cross_val_score,\n",
    "    cross_validate,\n",
    "    train_test_split,\n",
    ")\n",
    "from sklearn.pipeline import Pipeline, make_pipeline\n",
    "from sklearn.preprocessing import (\n",
    "    MinMaxScaler,\n",
    "    OneHotEncoder,\n",
    "    OrdinalEncoder,\n",
    "    StandardScaler,\n",
    ")\n",
    "\n",
    "plt.rcParams[\"font.size\"] = 16"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "## Learning objectives\n",
    "\n",
    "- When communicating about applied ML, tailor an explanation to the intended audience.\n",
    "- Apply best practices of technical communication, such as bottom-up explanations and reader-centric writing.\n",
    "- Given an ML problem, analyze the decision being made and the objectives.\n",
    "- Avoid the pitfall of thinking about ML as coding in isolation; build the habit of relating your work to the surrounding context and stakeholders.\n",
    "- Interpret a confidence score or credence, e.g. what does it mean to be 5% confident that a statement is true.\n",
    "- Maintain a healthy skepticism of `predict_proba` scores and their possible interpretation as credences.\n",
    "- Be careful and precise when communicating confidence to stakeholders in an ML project.\n",
    "- Identify misleading visualizations."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "## Attribution\n",
    "\n",
    "- The first part of this lecture is adapted from [DSCI 542](https://github.com/UBC-MDS/DSCI_542_comm-arg), created by [David Laing](https://davidklaing.com/).\n",
    "- The visualization component of this lecture benefitted from discussions with [Firas Moosvi](http://firas.moosvi.com/) about his course, [DSCI 531](https://github.com/UBC-MDS/DSCI_531_viz-1)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## Why should we care about effective communication?\n",
    "\n",
    "- Most ML practitioners work in an organization with >1 people.\n",
    "- There will very likely be stakeholders other than yourself.\n",
    "- Those people need to understand what you're doing because:\n",
    "  - their state of mind may change the way you do things (see below)\n",
    "  - your state of mind may change the way they do things (interpreting your results)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "ML suffers from some particular communication issues:\n",
    "- overstating one's results / unable to articulate the limitations\n",
    "- unable to explain the predictions\n",
    "- Can we trust test error?\n",
    "- Why did CatBoost make that prediction?\n",
    "- What does it mean if `predict_proba` outputs 0.9?\n",
    "\n",
    "These issues are there because these things are actually very hard to explain!"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## Activity 1: explaining `GridSearchCV`"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "-"
    }
   },
   "source": [
    "Below are two possible explanations of `GridSearchCV` pitched to different audiences. Read them both and then follow the instructions at the end."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Explanation 1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Machine learning algorithms, like an airplane's cockpit, typically involve a bunch of knobs and switches that need to be set.\n",
    "\n",
    "![](https://i.pinimg.com/236x/ea/43/f3/ea43f3c7f3a8c92d884ce012c77628fd--cockpit-gauges.jpg)\n",
    "\n",
    "For example, check out the documentation of the popular random forest algorithm [here](https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.RandomForestClassifier.html). Here's a list of the function arguments, along with their default values (from the documentation):\n",
    "\n",
    "> class sklearn.ensemble.RandomForestClassifier(n_estimators=100, criterion='gini', max_depth=None, min_samples_split=2, min_samples_leaf=1, min_weight_fraction_leaf=0.0, max_features='auto', max_leaf_nodes=None, min_impurity_decrease=0.0, min_impurity_split=None, bootstrap=True, oob_score=False, n_jobs=None, random_state=None, verbose=0, warm_start=False, class_weight=None, ccp_alpha=0.0, max_samples=None)\n",
    "\n",
    "Holy cow, that's a lot of knobs and switches! As a machine learning practitioner, how am I supposed to choose `n_estimators`? Should I leave it at the default of 100? Or try 1000? What about `criterion` or `class_weight` for that matter? Should I trust the defaults?\n",
    "\n",
    "Enter [GridSearchCV](https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.GridSearchCV.html) to save the day. The general strategy here is the choose the settings that perform best on the specific task of interest. So I can't say `n_estimators=100` is better than `n_estimators=1000` without knowing what problem I'm working on. For a specific problem, you usually have a numerical score that measures performance. `GridSearchCV` is part of the popular [scikit-learn](https://scikit-learn.org/) Python machine learning library. It works by searching over various settings and tells you which one worked best on your problem. \n",
    "\n",
    "The \"grid\" in \"grid search\" comes from the fact that tries all possible combinations on a grid. For example, if you want it to consider setting `n_estimators` to 100, 150 or 200, and you want it to consider setting `criterion` to `'gini'` or `'entropy'`, then it will search over all 6 possible combinations in a grid of 3 possible values by 2 possible values: \n",
    "\n",
    "|                    | `criterion='gini'` | `criterion='entropy'` |\n",
    "|----------------------|--------|---------|\n",
    "| `n_estimators=100` |    1     |    2     |\n",
    "| `n_estimators=150` |    3     |    4     |\n",
    "| `n_estimators=200` |    5     |    6     |\n",
    "\n",
    "Here is a code sample that uses `GridSearchCV` to select from the 6 options we just mentioned. The problem being solved is classifying images of handwritten digits into the 10 digit categories (0-9). I chose this because the dataset is conveniently built in to scikit-learn:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'criterion': 'gini', 'n_estimators': 100}"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# imports\n",
    "from sklearn import datasets\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "\n",
    "# load a dataset\n",
    "data = datasets.load_digits()\n",
    "X = data[\"data\"]\n",
    "y = data[\"target\"]\n",
    "\n",
    "# set up the grid search\n",
    "grid_search = GridSearchCV(\n",
    "    RandomForestClassifier(random_state=123),\n",
    "    param_grid={\"n_estimators\": [100, 150, 200], \"criterion\": [\"gini\", \"entropy\"]},\n",
    ")\n",
    "\n",
    "# run the grid search\n",
    "grid_search.fit(X, y)\n",
    "grid_search.best_params_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As we can see from the output above, the grid search selected `criterion='gini', n_estimators=100`, which was one of our 6 options above (specifically Option 1).\n",
    "\n",
    "By the way, these \"knobs\" we've been setting are called [_hyperparameters_](https://en.wikipedia.org/wiki/Hyperparameter_(machine_learning) and the process of setting these hyperparameters automatically is called [_hyperparameter optimization_](https://en.wikipedia.org/wiki/Hyperparameter_optimization) or _hyperparameter tuning_."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "~400 words, not including code."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "<br><br><br><br><br><br>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Explanation 2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "-"
    }
   },
   "source": [
    "https://medium.com/datadriveninvestor/an-introduction-to-grid-search-ff57adcc0998\n",
    "\n",
    "~400 words, not including code."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "<br><br><br><br><br><br>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Discussion questions:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- What do you like about each explanation?\n",
    "- What do you dislike about each explanation?\n",
    "- What do you think is the intended audience for each explanation?\n",
    "- Which explanation do you think is more effective overall for someone on Day 1 of CPSC 330?\n",
    "- Each explanation has an image. Which one is more effective? What are the pros/cons?\n",
    "- Each explanation has some sample code. Which one is more effective? What are the pros/cons?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "After you're done reading, take ~5 min to consider the discussion questions above. Paste your answer to **at least one** of the above questions in the [Google doc](https://docs.google.com/document/d/1k0D_DLDcbPLKwAClSr6J_0t32SSUySPSjV4F0tytZ6c/edit?usp=sharing) under the appropriate question heading."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## Principles of good explanations"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Concepts *then* labels, not the other way around\n",
    "\n",
    "The first explanation start with an analogy for the concept (and the label is left until the very end):\n",
    "\n",
    "> Machine learning algorithms, like an airplane's cockpit, typically involve a bunch of knobs and switches that need to be set.\n",
    "\n",
    "In the second explanation, the first sentence is wasted on anyone who doesn't already know what \"hyperparameter tuning\" means:\n",
    "\n",
    "> Grid search is the process of performing hyper parameter tuning in order to determine the optimal values for a given model. \n",
    "\n",
    "The effectiveness of these different statements depend on your audience. \n",
    "\n",
    "See [this video](https://twitter.com/ProfFeynman/status/899963856549625858?s=20): \"I learned very early the difference between knowing the name of something and knowing something.\" -Richard Feynman."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Bottom-up explanations"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The [Curse of Knowledge](https://en.wikipedia.org/wiki/Curse_of_knowledge) leads to *top-down* explanations:\n",
    "\n",
    "![](img/top_down.png)\n",
    "\n",
    "- When you know something well, you think about things in the context of all your knowledge. \n",
    "- Those lacking the context, or frame of mind, cannot easily understand. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    },
    "tags": []
   },
   "source": [
    "There is another way: *bottom-up* explanations:\n",
    "\n",
    "![](img/bottom_up.png)\n",
    "\n",
    "When you're brand new to a concept, you benefit from analogies, concrete examples and familiar patterns.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### New ideas in small chunks\n",
    "\n",
    "The first explanation has a hidden **conceptual skeleton**:\n",
    "\n",
    "1. The concept of setting a bunch of values.\n",
    "2. Random forest example.\n",
    "3. The problem / pain point.\n",
    "4. The solution.\n",
    "5. How it works - high level.\n",
    "6. How it works - written example.\n",
    "7. How it works - code example.\n",
    "8. The name of what we were discussing all this time."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Reuse your running examples\n",
    "\n",
    "Effective explanations often use the same example throughout the text and code. This helps readers follow the line of reasoning."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Approach from all angles\n",
    "\n",
    "When we're trying to draw mental boundaries around a concept, it's helpful to see **examples on all sides** of those boundaries. If we were writing a longer explanation, it might have been better to show more, e.g.\n",
    "\n",
    "- Performance **with** and **without** hyperparameter tuning.\n",
    "- **Other types** of hyperparameter tuning (e.g. `RandomizedSearchCV`)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### When experimenting, ***show the results*** asap\n",
    "\n",
    "The first explanation shows the output of the code, whereas the second does not. This is easy to do and makes a big difference."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Interesting to you != useful to the reader (aka it's not about you)\n",
    "\n",
    "Here is something which was deleted from the explanation:\n",
    "\n",
    "> Some hyperparameters, like `n_estimators` are numeric. **Numeric** hyperparameters are like the **knobs** in the cockpit: you can tune them **continuously**. `n_estimators` is numeric. **Categorical** hyperparameters are like the **switches** in the cockpit: they can take on (two or more) **distinct values**. `criterion` is categorical.\n",
    "\n",
    "It's a very **elegant analogy**! But is it **helpful?**\n",
    "\n",
    "And furthermore, what is my hidden motivation for wanting to include it? Elegance, art, and the pursuit of higher beauty? Or _making myself look smart_? So maybe another name for this principle could be **It's not about you.**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    },
    "tags": []
   },
   "source": [
    "## ML and decision-making"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- There is often a wide gap between what people care about and what ML can do.\n",
    "- To understand what ML can do, let's think about what **decisions** will be made using ML. \n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    },
    "tags": []
   },
   "source": [
    "#### Decisions involve a few key pieces\n",
    "\n",
    "- The **decision variable**: the variable that is manipulated through the decision.\n",
    "  - E.g., how much should I sell my house for? (numeric)\n",
    "  - E.g., should I sell my house? (categorical)\n",
    "- The decision-maker's **objectives**: the variables that the decision-maker **ultimately cares about**, and wishes to manipulate indirectly through the decision variable.\n",
    "  - E.g., my total profit, time to sale, etc.\n",
    "- The **context**: the variables that mediate the relationship between the decision variable and the objectives.\n",
    "  - E.g., the housing market, cost of marketing it, my timeline, etc."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "#### How does this inform you as an ML practitioner?\n",
    "\n",
    "Questions you have to answer:\n",
    "\n",
    "- Who is the decision maker?\n",
    "- What are their objectives?\n",
    "- What are their alternatives?\n",
    "- What is their context?\n",
    "- What data do I need?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## Break\n",
    "\n",
    "![](img/eva-coffee.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "<br><br>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## Confidence and `predict_proba`"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "- What does it mean to be \"confident\" in your results?\n",
    "- When you perform analysis, you are **responsible for many judgment** calls.\n",
    "- [Your results will be different than others'](https://fivethirtyeight.com/features/science-isnt-broken/#part1).\n",
    "- As you make these judgments and start to form conclusions, how can you recognize your own uncertainties about the data so that you can communicate confidently?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### What does this mean for us, when we're trying to make claims about our data?\n",
    "\n",
    "Let's **imagine** that the following claim **is true**:\n",
    "\n",
    "> Vancouver has the highest cost of living of all cities in Canada.\n",
    "\n",
    "Now let's consider a few beliefs we could hold:\n",
    "\n",
    "1. Vancouver has the highest cost of living of all cities in Canada. **I am 95% sure of this.** \n",
    "2. Vancouver has the highest cost of living of all cities in Canada. **I am 55% sure of this.** \n",
    "\n",
    "The part is bold is called a [credence](https://en.wikipedia.org/wiki/Credence_(statistics)). Which belief is better?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "But **what if it's actually Toronto** that has the highest cost of living in Canada?\n",
    "\n",
    "1. Vancouver has the highest cost of living of all cities in Canada. **I am 95% sure of this.** \n",
    "2. Vancouver has the highest cost of living of all cities in Canada. **I am 55% sure of this.** \n",
    "\n",
    "Which belief is better now?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "**Conclusion**: \n",
    "- We don't just want to be right.\n",
    "- We want to be ***confident* when we're right** and ***hesitant* when we're wrong.**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### What does credence mean in practical terms?\n",
    "\n",
    "One of two things:\n",
    "\n",
    "* **I would accept a bet at these odds.** 99% sure means, \"For the chance of winning \\\\$1, I would bet \\\\$99 that I'm right about this.\" 75% sure means, \"For the chance of winning \\\\$25, I would bet \\\\$75 that I'm right about this.\"\n",
    "* **Long-run frequency of correctness.** 99% sure means, \"For every 100 predictions I make at this level of confidence, I would expect only 1 of them to be incorrect.\" 75% sure means, \"For every 100 predictions I make at this level of confidence, I would expect about 25 of them to be incorrect.\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### It's easy enough to evaluate how good we are at being right...\n",
    "\n",
    "But if we want to **evaluate** *how good we are at knowing how right we are*?\n",
    "\n",
    "We would need to keep of not just the **correctness of our predictions**, but also our **confidence in those predictions**.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### What does this have to do with applied ML?\n",
    "\n",
    "- What if you `predict` that a credit card transaction is fraudulent?\n",
    "  - We probably want `predict_proba` a lot of the time.\n",
    "- What if `predict_proba` is 0.95 in that case?\n",
    "  - How confident are YOU?\n",
    "- What if you forecast that avocado prices will go up next week? \n",
    "  - How confident are you there?\n",
    "- Or what if you predict a house price to be \\\\$800k? \n",
    "  - That is not even a true/false statement."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Loss functions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "When you call `fit` for `LogisticRegression` it has these same **preferences**:\n",
    "\n",
    "correct and confident $>$ correct and hesitant $>$ incorrect and hesitant $>$ incorrect and confident."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "from sklearn.metrics import log_loss"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- This is a \"loss\" or \"error\" function like *mean squared error*, so **lower values are better**.\n",
    "- When you call `fit` it tries to minimize this metric."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "The [sklearn documentation on log_loss](https://scikit-learn.org/stable/modules/model_evaluation.html#log-loss) reads:\n",
    "> The `log_loss` function computes log loss given a list of ground-truth labels and a probability matrix, as returned by an estimator’s `predict_proba` method.\n",
    "\n",
    "<br>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "**Example**\n",
    "\n",
    "Let's consider **one sample** for which we have the **ground-truth** value `0`.\n",
    "- You can specify this as `y_true=[0]`. For more values, the list would be longer, of course.\n",
    "\n",
    "Also, let's say the estimator’s `predict_proba` method has returned `[0.95, 0.05]` for this sample.\n",
    "- You can specify this as `y_pred=[[0.95, 0.05]]`. \n",
    "- Equivalently, you can just specify the second value as `y_pred=[0.05]`.\n",
    "\n",
    "Finally, the call to `log_loss` function can be made as follows, where `labels` specifies all possible values for `y`:\n",
    "- `log_loss(y_true=[0], y_pred=[[0.95, 0.05]], labels=(0, 1))`"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "<br><br>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "Let's see some example cases for **correctness** and **credence**.<br>\n",
    "Assume the ground-truth is `0` in all the following examples."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "- example 1: prediction `0` is correct with 95% confident:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.05129329438755058"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "log_loss(y_true=[0], y_pred=[[0.95, 0.05]], labels=(0, 1))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "- example 2: prediction `0` is correct with 55% confident:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.5978370007556204"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "log_loss(y_true=[0], y_pred=[[0.55, 0.45]], labels=(0, 1))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "- example 3: prediction `0` is incorrect with 55% confident:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.7985076962177716"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "log_loss(y_true=[0], y_pred=[[0.45, 0.55]], labels=(0, 1))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "- example 4: prediction `0` is incorrect with 95% confident:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2.995732273553991"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "log_loss(y_true=[0], y_pred=[[0.05, 0.95]], labels=(0, 1))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "<br>\n",
    "Let us see this effect on a graph:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "grid = np.linspace(0, 1, 1000)\n",
    "data = {\n",
    "    \"Ground-truth 0\": [log_loss(y_true=[0], y_pred=[[1-g, g]], labels=(0, 1)) for g in grid],\n",
    "    \"Ground-truth 1\": [log_loss(y_true=[1], y_pred=[[1-g, g]], labels=(0, 1)) for g in grid],\n",
    "}\n",
    "pd.DataFrame(data, index=grid).plot()\n",
    "plt.xlabel(\"Confidence in predicting 1\")\n",
    "plt.ylabel(\"Loss (lower is better)\");\n",
    "plt.ylim(0, 5);"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For ground-truth is `1` and prediction is `1`:\n",
    "- Your loss goes to 0 as you approach 100% confidence in the **correct** answer (`1`).\n",
    "- Your loss goes to infinity as you approach 0% confidence in the **correct** answer (`1`).\n",
    "\n",
    "For ground-truth is `0` but prediction is `1`:\n",
    "- Your loss goes to 0 as you approach 0% confidence in the **incorrect** answer (`1`).\n",
    "- Your loss goes to infinity as you approach 100% confidence in the **incorrect** answer (`1`)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "(Optional) See also the very related [How to assign partial credit on an exam of true-false questions?](https://terrytao.wordpress.com/2016/06/01/how-to-assign-partial-credit-on-an-exam-of-true-false-questions/)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The real `LogisticRegression` is averaging this score over all training examples."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "\n",
    "Some nice examples: \n",
    "\n",
    "- [Scott Alexander](https://slatestarcodex.com/2019/01/22/2018-predictions-calibration-results/)\n",
    "  - Look at how the plot starts at 50%. That is because being 40% confident of \"X\" is the same as being 60% confident of \"not X\".\n",
    "- [Good Judgment Project](https://www.gjopen.com/)\n",
    "\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## Activity 2: Credence (time permitting)\n",
    "\n",
    "- Take a few minutes and assign credences or values to the claims below, in the [Google doc](https://docs.google.com/document/d/1k0D_DLDcbPLKwAClSr6J_0t32SSUySPSjV4F0tytZ6c/edit?usp=sharing).  Afterwards, we'll discuss.\n",
    "- **Do not search the answers; the point of the exercise is to evaluate how good we are at guessing.** \n",
    "- Try not to be influenced by other peoples' answers! Better to pick your answers before going to the Google Doc.\n",
    "\n",
    "1. I am __ % sure that the world's longest river is between 6000km and 8000km. \n",
    "2. I am __ % sure that there is 4 to 8 liters of blood in an average adult human body. \n",
    "3. I am 99% sure that the world's tallest tree taller than __ m.\n",
    "4. I am 90% sure that the world's tallest tree taller than __ m.\n",
    "5. I am 50% sure that the world's tallest tree taller than __ m.\n",
    "\n",
    "> NOTE: 100% means you are completely sure the statement is true, 0% means you are completely sure the statement is false."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "<br><br>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## Visualizing your results"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "- Very powerful but at the same time can be misleading if not done properly. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "Pre-viewing review from [Calling BS visualization videos](https://www.youtube.com/watch?v=T-5aLbNeGo0&list=PLPnZfvKID1Sje5jWxt-4CSZD7bUI4gSPS&index=30&t=0s):\n",
    "\n",
    "- Dataviz in the popular media.\n",
    "  - e.g. [modern NYT](https://youtu.be/T-5aLbNeGo0?t=367)\n",
    "- Misleading axes.\n",
    "  - e.g. [vaccines](https://youtu.be/9pNWVMxaFuM?t=299)\n",
    "- Manipulating bin sizes.\n",
    "  - e.g. [tax dollars](https://youtu.be/zAg1wsYfwsM?t=196)\n",
    "- Dataviz ducks.\n",
    "  - e.g. [drinking water](https://youtu.be/rmii1hfP6d4?t=169)\n",
    "  - \"look how clever we are about design\" -> making it about me instead of about you (see last class)\n",
    "- Glass slippers.\n",
    "  - e.g. [internet marketing tree](https://youtu.be/59teS0SUHtI?t=285)\n",
    "- The principle of proportional ink.\n",
    "  - e.g. [most read books](https://youtu.be/oNhusd3xFC4?t=147)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "- [Demo of cleaning up a plot](https://www.darkhorseanalytics.com/blog/data-looks-better-naked)\n",
    "- [Principle of proportional ink](https://serialmentor.com/dataviz/proportional-ink.html) from \"Fundamentals of Data Visualization\" by Claus O. Wilke:\n",
    "> The sizes of shaded areas in a visualization need to be proportional to the data values they represent."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Dataset\n",
    "\n",
    "We'll be using [Kaggle House Prices dataset](https://www.kaggle.com/c/home-data-for-ml-course/), which we used in lecture 10. As usual, to run this notebook you'll need to download the data. For this dataset, train and test have already been separated. We'll be working with the train portion. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Id</th>\n",
       "      <th>MSSubClass</th>\n",
       "      <th>MSZoning</th>\n",
       "      <th>LotFrontage</th>\n",
       "      <th>LotArea</th>\n",
       "      <th>Street</th>\n",
       "      <th>Alley</th>\n",
       "      <th>LotShape</th>\n",
       "      <th>LandContour</th>\n",
       "      <th>Utilities</th>\n",
       "      <th>...</th>\n",
       "      <th>PoolArea</th>\n",
       "      <th>PoolQC</th>\n",
       "      <th>Fence</th>\n",
       "      <th>MiscFeature</th>\n",
       "      <th>MiscVal</th>\n",
       "      <th>MoSold</th>\n",
       "      <th>YrSold</th>\n",
       "      <th>SaleType</th>\n",
       "      <th>SaleCondition</th>\n",
       "      <th>SalePrice</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>302</th>\n",
       "      <td>303</td>\n",
       "      <td>20</td>\n",
       "      <td>RL</td>\n",
       "      <td>118.0</td>\n",
       "      <td>13704</td>\n",
       "      <td>Pave</td>\n",
       "      <td>NaN</td>\n",
       "      <td>IR1</td>\n",
       "      <td>Lvl</td>\n",
       "      <td>AllPub</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2006</td>\n",
       "      <td>WD</td>\n",
       "      <td>Normal</td>\n",
       "      <td>205000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>767</th>\n",
       "      <td>768</td>\n",
       "      <td>50</td>\n",
       "      <td>RL</td>\n",
       "      <td>75.0</td>\n",
       "      <td>12508</td>\n",
       "      <td>Pave</td>\n",
       "      <td>NaN</td>\n",
       "      <td>IR1</td>\n",
       "      <td>Lvl</td>\n",
       "      <td>AllPub</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Shed</td>\n",
       "      <td>1300</td>\n",
       "      <td>7</td>\n",
       "      <td>2008</td>\n",
       "      <td>WD</td>\n",
       "      <td>Normal</td>\n",
       "      <td>160000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>429</th>\n",
       "      <td>430</td>\n",
       "      <td>20</td>\n",
       "      <td>RL</td>\n",
       "      <td>130.0</td>\n",
       "      <td>11457</td>\n",
       "      <td>Pave</td>\n",
       "      <td>NaN</td>\n",
       "      <td>IR1</td>\n",
       "      <td>Lvl</td>\n",
       "      <td>AllPub</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>2009</td>\n",
       "      <td>WD</td>\n",
       "      <td>Normal</td>\n",
       "      <td>175000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1139</th>\n",
       "      <td>1140</td>\n",
       "      <td>30</td>\n",
       "      <td>RL</td>\n",
       "      <td>98.0</td>\n",
       "      <td>8731</td>\n",
       "      <td>Pave</td>\n",
       "      <td>NaN</td>\n",
       "      <td>IR1</td>\n",
       "      <td>Lvl</td>\n",
       "      <td>AllPub</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0</td>\n",
       "      <td>5</td>\n",
       "      <td>2007</td>\n",
       "      <td>WD</td>\n",
       "      <td>Normal</td>\n",
       "      <td>144000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>558</th>\n",
       "      <td>559</td>\n",
       "      <td>60</td>\n",
       "      <td>RL</td>\n",
       "      <td>57.0</td>\n",
       "      <td>21872</td>\n",
       "      <td>Pave</td>\n",
       "      <td>NaN</td>\n",
       "      <td>IR2</td>\n",
       "      <td>HLS</td>\n",
       "      <td>AllPub</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0</td>\n",
       "      <td>8</td>\n",
       "      <td>2008</td>\n",
       "      <td>WD</td>\n",
       "      <td>Normal</td>\n",
       "      <td>175000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 81 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "        Id  MSSubClass MSZoning  LotFrontage  LotArea Street Alley LotShape  \\\n",
       "302    303          20       RL        118.0    13704   Pave   NaN      IR1   \n",
       "767    768          50       RL         75.0    12508   Pave   NaN      IR1   \n",
       "429    430          20       RL        130.0    11457   Pave   NaN      IR1   \n",
       "1139  1140          30       RL         98.0     8731   Pave   NaN      IR1   \n",
       "558    559          60       RL         57.0    21872   Pave   NaN      IR2   \n",
       "\n",
       "     LandContour Utilities  ... PoolArea PoolQC Fence MiscFeature MiscVal  \\\n",
       "302          Lvl    AllPub  ...        0    NaN   NaN         NaN       0   \n",
       "767          Lvl    AllPub  ...        0    NaN   NaN        Shed    1300   \n",
       "429          Lvl    AllPub  ...        0    NaN   NaN         NaN       0   \n",
       "1139         Lvl    AllPub  ...        0    NaN   NaN         NaN       0   \n",
       "558          HLS    AllPub  ...        0    NaN   NaN         NaN       0   \n",
       "\n",
       "     MoSold YrSold  SaleType  SaleCondition  SalePrice  \n",
       "302       1   2006        WD         Normal     205000  \n",
       "767       7   2008        WD         Normal     160000  \n",
       "429       3   2009        WD         Normal     175000  \n",
       "1139      5   2007        WD         Normal     144000  \n",
       "558       8   2008        WD         Normal     175000  \n",
       "\n",
       "[5 rows x 81 columns]"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.read_csv(\"data/housing-kaggle/train.csv\")\n",
    "train_df, test_df = train_test_split(df, test_size=0.10, random_state=123)\n",
    "train_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1314, 81)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Let's separate `X` and `y`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "slideshow": {
     "slide_type": "-"
    }
   },
   "outputs": [],
   "source": [
    "X_train = train_df.drop(columns=[\"SalePrice\"])\n",
    "y_train = train_df[\"SalePrice\"]\n",
    "\n",
    "X_test = test_df.drop(columns=[\"SalePrice\"])\n",
    "y_test = test_df[\"SalePrice\"]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Feature types \n",
    "\n",
    "- We have mixed feature types and a bunch of missing values. \n",
    "- Now, let's identify feature types and transformations. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "drop_features = [\"Id\"]\n",
    "numeric_features = [\n",
    "    \"BedroomAbvGr\",\n",
    "    \"KitchenAbvGr\",\n",
    "    \"LotFrontage\",\n",
    "    \"LotArea\",\n",
    "    \"OverallQual\",\n",
    "    \"OverallCond\",\n",
    "    \"YearBuilt\",\n",
    "    \"YearRemodAdd\",\n",
    "    \"MasVnrArea\",\n",
    "    \"BsmtFinSF1\",\n",
    "    \"BsmtFinSF2\",\n",
    "    \"BsmtUnfSF\",\n",
    "    \"TotalBsmtSF\",\n",
    "    \"1stFlrSF\",\n",
    "    \"2ndFlrSF\",\n",
    "    \"LowQualFinSF\",\n",
    "    \"GrLivArea\",\n",
    "    \"BsmtFullBath\",\n",
    "    \"BsmtHalfBath\",\n",
    "    \"FullBath\",\n",
    "    \"HalfBath\",\n",
    "    \"TotRmsAbvGrd\",\n",
    "    \"Fireplaces\",\n",
    "    \"GarageYrBlt\",\n",
    "    \"GarageCars\",\n",
    "    \"GarageArea\",\n",
    "    \"WoodDeckSF\",\n",
    "    \"OpenPorchSF\",\n",
    "    \"EnclosedPorch\",\n",
    "    \"3SsnPorch\",\n",
    "    \"ScreenPorch\",\n",
    "    \"PoolArea\",\n",
    "    \"MiscVal\",\n",
    "    \"YrSold\",\n",
    "]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[['Po', 'Fa', 'TA', 'Gd', 'Ex'],\n",
       " ['Po', 'Fa', 'TA', 'Gd', 'Ex'],\n",
       " ['Po', 'Fa', 'TA', 'Gd', 'Ex'],\n",
       " ['Po', 'Fa', 'TA', 'Gd', 'Ex'],\n",
       " ['Po', 'Fa', 'TA', 'Gd', 'Ex'],\n",
       " ['Po', 'Fa', 'TA', 'Gd', 'Ex'],\n",
       " ['Po', 'Fa', 'TA', 'Gd', 'Ex'],\n",
       " ['Po', 'Fa', 'TA', 'Gd', 'Ex'],\n",
       " ['Po', 'Fa', 'TA', 'Gd', 'Ex'],\n",
       " ['Po', 'Fa', 'TA', 'Gd', 'Ex']]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ordinal_features_reg = [\n",
    "    \"ExterQual\",\n",
    "    \"ExterCond\",\n",
    "    \"BsmtQual\",\n",
    "    \"BsmtCond\",\n",
    "    \"HeatingQC\",\n",
    "    \"KitchenQual\",\n",
    "    \"FireplaceQu\",\n",
    "    \"GarageQual\",\n",
    "    \"GarageCond\",\n",
    "    \"PoolQC\",\n",
    "]\n",
    "ordering = [\n",
    "    \"Po\",\n",
    "    \"Fa\",\n",
    "    \"TA\",\n",
    "    \"Gd\",\n",
    "    \"Ex\",\n",
    "]  # if N/A it will just impute something, per below\n",
    "ordering_ordinal_reg = [ordering] * len(ordinal_features_reg)\n",
    "ordering_ordinal_reg"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "ordinal_features_oth = [\n",
    "    \"BsmtExposure\",\n",
    "    \"BsmtFinType1\",\n",
    "    \"BsmtFinType2\",\n",
    "    \"Functional\",\n",
    "    \"Fence\",\n",
    "]\n",
    "ordering_ordinal_oth = [\n",
    "    [\"NA\", \"No\", \"Mn\", \"Av\", \"Gd\"],\n",
    "    [\"NA\", \"Unf\", \"LwQ\", \"Rec\", \"BLQ\", \"ALQ\", \"GLQ\"],\n",
    "    [\"NA\", \"Unf\", \"LwQ\", \"Rec\", \"BLQ\", \"ALQ\", \"GLQ\"],\n",
    "    [\"Sal\", \"Sev\", \"Maj2\", \"Maj1\", \"Mod\", \"Min2\", \"Min1\", \"Typ\"],\n",
    "    [\"NA\", \"MnWw\", \"GdWo\", \"MnPrv\", \"GdPrv\"],\n",
    "]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "The remaining features are categorical features."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['Exterior2nd',\n",
       " 'Electrical',\n",
       " 'SaleType',\n",
       " 'RoofStyle',\n",
       " 'CentralAir',\n",
       " 'Alley',\n",
       " 'LotShape',\n",
       " 'RoofMatl',\n",
       " 'MasVnrType',\n",
       " 'Exterior1st',\n",
       " 'GarageFinish',\n",
       " 'Street',\n",
       " 'BldgType',\n",
       " 'Condition1',\n",
       " 'HouseStyle',\n",
       " 'LandContour',\n",
       " 'Utilities',\n",
       " 'SaleCondition',\n",
       " 'PavedDrive',\n",
       " 'LandSlope',\n",
       " 'Heating',\n",
       " 'Foundation',\n",
       " 'GarageType',\n",
       " 'Neighborhood',\n",
       " 'Condition2',\n",
       " 'MiscFeature',\n",
       " 'MSSubClass',\n",
       " 'MoSold',\n",
       " 'MSZoning',\n",
       " 'LotConfig']"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "categorical_features = list(\n",
    "    set(X_train.columns)\n",
    "    - set(numeric_features)\n",
    "    - set(ordinal_features_reg)\n",
    "    - set(ordinal_features_oth)\n",
    "    - set(drop_features)\n",
    ")\n",
    "categorical_features"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Applying feature transformations\n",
    "\n",
    "- Since we have mixed feature types, let's use `ColumnTransformer` to apply different transformations on different features types.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "from sklearn.compose import ColumnTransformer, make_column_transformer\n",
    "\n",
    "numeric_transformer = make_pipeline(SimpleImputer(strategy=\"median\"), StandardScaler())\n",
    "ordinal_transformer_reg = make_pipeline(\n",
    "    SimpleImputer(strategy=\"most_frequent\"),\n",
    "    OrdinalEncoder(categories=ordering_ordinal_reg),\n",
    ")\n",
    "\n",
    "ordinal_transformer_oth = make_pipeline(\n",
    "    SimpleImputer(strategy=\"most_frequent\"),\n",
    "    OrdinalEncoder(categories=ordering_ordinal_oth),\n",
    ")\n",
    "\n",
    "categorical_transformer = make_pipeline(\n",
    "    SimpleImputer(strategy=\"constant\", fill_value=\"missing\"),\n",
    "    OneHotEncoder(handle_unknown=\"ignore\", sparse=False),\n",
    ")\n",
    "\n",
    "preprocessor = make_column_transformer(\n",
    "    (\"drop\", drop_features),\n",
    "    (numeric_transformer, numeric_features),\n",
    "    (ordinal_transformer_reg, ordinal_features_reg),\n",
    "    (ordinal_transformer_oth, ordinal_features_oth),\n",
    "    (categorical_transformer, categorical_features),\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Examining the preprocessed data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "ColumnTransformer(transformers=[('drop', 'drop', ['Id']),\n",
       "                                ('pipeline-1',\n",
       "                                 Pipeline(steps=[('simpleimputer',\n",
       "                                                  SimpleImputer(strategy='median')),\n",
       "                                                 ('standardscaler',\n",
       "                                                  StandardScaler())]),\n",
       "                                 ['BedroomAbvGr', 'KitchenAbvGr', 'LotFrontage',\n",
       "                                  'LotArea', 'OverallQual', 'OverallCond',\n",
       "                                  'YearBuilt', 'YearRemodAdd', 'MasVnrArea',\n",
       "                                  'BsmtFinSF1', 'BsmtFinSF2', 'BsmtUnfSF',\n",
       "                                  'TotalBsmtSF', '...\n",
       "                                 ['Exterior2nd', 'Electrical', 'SaleType',\n",
       "                                  'RoofStyle', 'CentralAir', 'Alley',\n",
       "                                  'LotShape', 'RoofMatl', 'MasVnrType',\n",
       "                                  'Exterior1st', 'GarageFinish', 'Street',\n",
       "                                  'BldgType', 'Condition1', 'HouseStyle',\n",
       "                                  'LandContour', 'Utilities', 'SaleCondition',\n",
       "                                  'PavedDrive', 'LandSlope', 'Heating',\n",
       "                                  'Foundation', 'GarageType', 'Neighborhood',\n",
       "                                  'Condition2', 'MiscFeature', 'MSSubClass',\n",
       "                                  'MoSold', 'MSZoning', 'LotConfig'])])"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "preprocessor.fit(X_train)\n",
    "# Calling fit to examine all the transformers."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "ohe_columns = list(\n",
    "    preprocessor.named_transformers_[\"pipeline-4\"]\n",
    "    .named_steps[\"onehotencoder\"]\n",
    "    .get_feature_names_out(categorical_features)\n",
    ")\n",
    "new_columns = (\n",
    "    numeric_features + ordinal_features_reg + ordinal_features_oth + ohe_columns\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "slideshow": {
     "slide_type": "-"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>BedroomAbvGr</th>\n",
       "      <th>KitchenAbvGr</th>\n",
       "      <th>LotFrontage</th>\n",
       "      <th>LotArea</th>\n",
       "      <th>OverallQual</th>\n",
       "      <th>OverallCond</th>\n",
       "      <th>YearBuilt</th>\n",
       "      <th>YearRemodAdd</th>\n",
       "      <th>MasVnrArea</th>\n",
       "      <th>BsmtFinSF1</th>\n",
       "      <th>...</th>\n",
       "      <th>MSZoning_C (all)</th>\n",
       "      <th>MSZoning_FV</th>\n",
       "      <th>MSZoning_RH</th>\n",
       "      <th>MSZoning_RL</th>\n",
       "      <th>MSZoning_RM</th>\n",
       "      <th>LotConfig_Corner</th>\n",
       "      <th>LotConfig_CulDSac</th>\n",
       "      <th>LotConfig_FR2</th>\n",
       "      <th>LotConfig_FR3</th>\n",
       "      <th>LotConfig_Inside</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>302</th>\n",
       "      <td>0.154795</td>\n",
       "      <td>-0.222647</td>\n",
       "      <td>2.312501</td>\n",
       "      <td>0.381428</td>\n",
       "      <td>0.663680</td>\n",
       "      <td>-0.512408</td>\n",
       "      <td>0.993969</td>\n",
       "      <td>0.840492</td>\n",
       "      <td>0.269972</td>\n",
       "      <td>-0.961498</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>767</th>\n",
       "      <td>1.372763</td>\n",
       "      <td>-0.222647</td>\n",
       "      <td>0.260890</td>\n",
       "      <td>0.248457</td>\n",
       "      <td>-0.054669</td>\n",
       "      <td>1.285467</td>\n",
       "      <td>-1.026793</td>\n",
       "      <td>0.016525</td>\n",
       "      <td>-0.573129</td>\n",
       "      <td>0.476092</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>429</th>\n",
       "      <td>0.154795</td>\n",
       "      <td>-0.222647</td>\n",
       "      <td>2.885044</td>\n",
       "      <td>0.131607</td>\n",
       "      <td>-0.054669</td>\n",
       "      <td>-0.512408</td>\n",
       "      <td>0.563314</td>\n",
       "      <td>0.161931</td>\n",
       "      <td>-0.573129</td>\n",
       "      <td>1.227559</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1139</th>\n",
       "      <td>0.154795</td>\n",
       "      <td>-0.222647</td>\n",
       "      <td>1.358264</td>\n",
       "      <td>-0.171468</td>\n",
       "      <td>-0.773017</td>\n",
       "      <td>-0.512408</td>\n",
       "      <td>-1.689338</td>\n",
       "      <td>-1.679877</td>\n",
       "      <td>-0.573129</td>\n",
       "      <td>0.443419</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>558</th>\n",
       "      <td>0.154795</td>\n",
       "      <td>-0.222647</td>\n",
       "      <td>-0.597924</td>\n",
       "      <td>1.289541</td>\n",
       "      <td>0.663680</td>\n",
       "      <td>-0.512408</td>\n",
       "      <td>0.828332</td>\n",
       "      <td>0.598149</td>\n",
       "      <td>-0.573129</td>\n",
       "      <td>0.354114</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 263 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      BedroomAbvGr  KitchenAbvGr  LotFrontage   LotArea  OverallQual  \\\n",
       "302       0.154795     -0.222647     2.312501  0.381428     0.663680   \n",
       "767       1.372763     -0.222647     0.260890  0.248457    -0.054669   \n",
       "429       0.154795     -0.222647     2.885044  0.131607    -0.054669   \n",
       "1139      0.154795     -0.222647     1.358264 -0.171468    -0.773017   \n",
       "558       0.154795     -0.222647    -0.597924  1.289541     0.663680   \n",
       "\n",
       "      OverallCond  YearBuilt  YearRemodAdd  MasVnrArea  BsmtFinSF1  ...  \\\n",
       "302     -0.512408   0.993969      0.840492    0.269972   -0.961498  ...   \n",
       "767      1.285467  -1.026793      0.016525   -0.573129    0.476092  ...   \n",
       "429     -0.512408   0.563314      0.161931   -0.573129    1.227559  ...   \n",
       "1139    -0.512408  -1.689338     -1.679877   -0.573129    0.443419  ...   \n",
       "558     -0.512408   0.828332      0.598149   -0.573129    0.354114  ...   \n",
       "\n",
       "      MSZoning_C (all)  MSZoning_FV  MSZoning_RH  MSZoning_RL  MSZoning_RM  \\\n",
       "302                0.0          0.0          0.0          1.0          0.0   \n",
       "767                0.0          0.0          0.0          1.0          0.0   \n",
       "429                0.0          0.0          0.0          1.0          0.0   \n",
       "1139               0.0          0.0          0.0          1.0          0.0   \n",
       "558                0.0          0.0          0.0          1.0          0.0   \n",
       "\n",
       "      LotConfig_Corner  LotConfig_CulDSac  LotConfig_FR2  LotConfig_FR3  \\\n",
       "302                1.0                0.0            0.0            0.0   \n",
       "767                0.0                0.0            0.0            0.0   \n",
       "429                1.0                0.0            0.0            0.0   \n",
       "1139               0.0                0.0            0.0            0.0   \n",
       "558                0.0                0.0            1.0            0.0   \n",
       "\n",
       "      LotConfig_Inside  \n",
       "302                0.0  \n",
       "767                1.0  \n",
       "429                0.0  \n",
       "1139               1.0  \n",
       "558                0.0  \n",
       "\n",
       "[5 rows x 263 columns]"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train_enc = pd.DataFrame(\n",
    "    preprocessor.transform(X_train), index=X_train.index, columns=new_columns\n",
    ")\n",
    "X_train_enc.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>BedroomAbvGr</th>\n",
       "      <th>KitchenAbvGr</th>\n",
       "      <th>LotFrontage</th>\n",
       "      <th>LotArea</th>\n",
       "      <th>OverallQual</th>\n",
       "      <th>OverallCond</th>\n",
       "      <th>YearBuilt</th>\n",
       "      <th>YearRemodAdd</th>\n",
       "      <th>MasVnrArea</th>\n",
       "      <th>BsmtFinSF1</th>\n",
       "      <th>...</th>\n",
       "      <th>MSZoning_C (all)</th>\n",
       "      <th>MSZoning_FV</th>\n",
       "      <th>MSZoning_RH</th>\n",
       "      <th>MSZoning_RL</th>\n",
       "      <th>MSZoning_RM</th>\n",
       "      <th>LotConfig_Corner</th>\n",
       "      <th>LotConfig_CulDSac</th>\n",
       "      <th>LotConfig_FR2</th>\n",
       "      <th>LotConfig_FR3</th>\n",
       "      <th>LotConfig_Inside</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>147</th>\n",
       "      <td>0.154795</td>\n",
       "      <td>-0.222647</td>\n",
       "      <td>-0.025381</td>\n",
       "      <td>-0.085415</td>\n",
       "      <td>0.663680</td>\n",
       "      <td>-0.512408</td>\n",
       "      <td>0.993969</td>\n",
       "      <td>0.792023</td>\n",
       "      <td>0.438592</td>\n",
       "      <td>-0.961498</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>676</th>\n",
       "      <td>1.372763</td>\n",
       "      <td>4.348569</td>\n",
       "      <td>-0.454788</td>\n",
       "      <td>-0.074853</td>\n",
       "      <td>-1.491366</td>\n",
       "      <td>-3.209221</td>\n",
       "      <td>-2.351883</td>\n",
       "      <td>-1.679877</td>\n",
       "      <td>-0.573129</td>\n",
       "      <td>-0.961498</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1304</th>\n",
       "      <td>0.154795</td>\n",
       "      <td>-0.222647</td>\n",
       "      <td>-1.790721</td>\n",
       "      <td>-0.768279</td>\n",
       "      <td>0.663680</td>\n",
       "      <td>-0.512408</td>\n",
       "      <td>1.093350</td>\n",
       "      <td>0.937429</td>\n",
       "      <td>0.084490</td>\n",
       "      <td>-0.961498</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1372</th>\n",
       "      <td>0.154795</td>\n",
       "      <td>-0.222647</td>\n",
       "      <td>0.260890</td>\n",
       "      <td>-0.058176</td>\n",
       "      <td>0.663680</td>\n",
       "      <td>0.386530</td>\n",
       "      <td>0.894587</td>\n",
       "      <td>0.646617</td>\n",
       "      <td>-0.573129</td>\n",
       "      <td>1.162214</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1427</th>\n",
       "      <td>0.154795</td>\n",
       "      <td>-0.222647</td>\n",
       "      <td>-0.454788</td>\n",
       "      <td>0.073016</td>\n",
       "      <td>-0.773017</td>\n",
       "      <td>0.386530</td>\n",
       "      <td>-0.861157</td>\n",
       "      <td>-1.679877</td>\n",
       "      <td>-0.573129</td>\n",
       "      <td>0.301838</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 263 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      BedroomAbvGr  KitchenAbvGr  LotFrontage   LotArea  OverallQual  \\\n",
       "147       0.154795     -0.222647    -0.025381 -0.085415     0.663680   \n",
       "676       1.372763      4.348569    -0.454788 -0.074853    -1.491366   \n",
       "1304      0.154795     -0.222647    -1.790721 -0.768279     0.663680   \n",
       "1372      0.154795     -0.222647     0.260890 -0.058176     0.663680   \n",
       "1427      0.154795     -0.222647    -0.454788  0.073016    -0.773017   \n",
       "\n",
       "      OverallCond  YearBuilt  YearRemodAdd  MasVnrArea  BsmtFinSF1  ...  \\\n",
       "147     -0.512408   0.993969      0.792023    0.438592   -0.961498  ...   \n",
       "676     -3.209221  -2.351883     -1.679877   -0.573129   -0.961498  ...   \n",
       "1304    -0.512408   1.093350      0.937429    0.084490   -0.961498  ...   \n",
       "1372     0.386530   0.894587      0.646617   -0.573129    1.162214  ...   \n",
       "1427     0.386530  -0.861157     -1.679877   -0.573129    0.301838  ...   \n",
       "\n",
       "      MSZoning_C (all)  MSZoning_FV  MSZoning_RH  MSZoning_RL  MSZoning_RM  \\\n",
       "147                0.0          0.0          0.0          1.0          0.0   \n",
       "676                0.0          0.0          0.0          0.0          1.0   \n",
       "1304               0.0          0.0          0.0          0.0          1.0   \n",
       "1372               0.0          0.0          0.0          1.0          0.0   \n",
       "1427               0.0          0.0          0.0          1.0          0.0   \n",
       "\n",
       "      LotConfig_Corner  LotConfig_CulDSac  LotConfig_FR2  LotConfig_FR3  \\\n",
       "147                0.0                1.0            0.0            0.0   \n",
       "676                0.0                0.0            0.0            0.0   \n",
       "1304               0.0                0.0            0.0            0.0   \n",
       "1372               1.0                0.0            0.0            0.0   \n",
       "1427               0.0                0.0            0.0            0.0   \n",
       "\n",
       "      LotConfig_Inside  \n",
       "147                0.0  \n",
       "676                1.0  \n",
       "1304               1.0  \n",
       "1372               0.0  \n",
       "1427               1.0  \n",
       "\n",
       "[5 rows x 263 columns]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_test_enc = pd.DataFrame(\n",
    "    preprocessor.transform(X_test), index=X_test.index, columns=new_columns\n",
    ")\n",
    "X_test_enc.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((1314, 80), (146, 80))"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape, X_test.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Training random forests and gradient boosted trees"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.ensemble import GradientBoostingRegressor"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's compare sklearn's `GradientBoostingRegressor` to `RandomForestRegressor` for different values of `n_estimators`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "n_estimators_values = [3, 10, 30, 100, 300]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3\n",
      "10\n",
      "30\n",
      "100\n",
      "300\n"
     ]
    }
   ],
   "source": [
    "score_rf_train = list()\n",
    "score_rf_test = list()\n",
    "score_gb_train = list()\n",
    "score_gb_test = list()\n",
    "\n",
    "for n_estimators in n_estimators_values:\n",
    "    print(n_estimators)\n",
    "    rf = TransformedTargetRegressor(\n",
    "        RandomForestRegressor(n_estimators=n_estimators, random_state=123),\n",
    "        func=np.log1p,\n",
    "        inverse_func=np.expm1,\n",
    "    )\n",
    "    rf.fit(X_train_enc, y_train)\n",
    "    score_rf_train.append(rf.score(X_train_enc, y_train))\n",
    "    score_rf_test.append(rf.score(X_test_enc, y_test))\n",
    "\n",
    "    gb = TransformedTargetRegressor(\n",
    "        GradientBoostingRegressor(n_estimators=n_estimators, random_state=123),\n",
    "        func=np.log1p,\n",
    "        inverse_func=np.expm1,\n",
    "    )\n",
    "    gb.fit(X_train_enc, y_train)\n",
    "    score_gb_train.append(gb.score(X_train_enc, y_train))\n",
    "    score_gb_test.append(gb.score(X_test_enc, y_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "Here is a **low-quality plot that is confusing** and perhaps downright misleading:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.subplot(2, 1, 1)\n",
    "plt.plot(n_estimators_values, score_rf_train, label=\"rf train\")\n",
    "plt.plot(n_estimators_values, score_rf_test, label=\"rf test\")\n",
    "plt.ylabel(\"$R^2$ score\")\n",
    "plt.legend()\n",
    "plt.subplot(2, 1, 2)\n",
    "plt.plot(n_estimators_values, score_gb_train, label=\"gb train\")\n",
    "plt.plot(n_estimators_values, score_gb_test, label=\"gb test\")\n",
    "plt.xlabel(\"n estimators\")\n",
    "plt.ylabel(\"$R^2$ score\")\n",
    "plt.legend()\n",
    "plt.tight_layout();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(n_estimators_values, score_rf_train, label=\"rf train\")\n",
    "plt.plot(n_estimators_values, score_rf_test, label=\"rf test\")\n",
    "\n",
    "plt.plot(n_estimators_values, score_gb_train, label=\"gb train\")\n",
    "plt.plot(n_estimators_values, score_gb_test, label=\"gb test\")\n",
    "plt.xlabel(\"n estimators\")\n",
    "plt.ylabel(\"$R^2$ score\")\n",
    "plt.legend();"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "Let's create some misleading visualizations\n",
    "\n",
    "- Create a visualization that makes RF look better than GB.\n",
    "- Create a visualization that makes GB look better than RF.\n",
    "- Create a visualization that makes RF and GB look equally good."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "#### Activity 3\n",
    "\n",
    "You can create your own misleading example and copy or screenshot it and paste it into the [Google Doc](https://docs.google.com/document/d/1k0D_DLDcbPLKwAClSr6J_0t32SSUySPSjV4F0tytZ6c/edit?usp=sharing).\n",
    "\n",
    "Here are some misleading plots. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### RF better than GB"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "nmax = 3\n",
    "\n",
    "# plt.plot(n_estimators_values[:nmax], score_rf_train[:nmax], label=\"rf train\")\n",
    "plt.plot(n_estimators_values[:nmax], score_rf_test[:nmax], label=\"rf test\")\n",
    "# plt.plot(n_estimators_values[:nmax], score_gb_train[:nmax], label=\"gb train\")\n",
    "plt.plot(n_estimators_values[:nmax], score_gb_test[:nmax], label=\"gb test\")\n",
    "plt.xlabel(\"n estimators\")\n",
    "plt.ylabel(\"$R^2$ score\")\n",
    "plt.legend()\n",
    "plt.title(\"for most values of n_estimators, RF is betteR!!\");"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### GB better than RF"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "nmin = 3\n",
    "\n",
    "# plt.plot(n_estimators_values[nmin:], score_rf_train[nmin:], label=\"rf train\")\n",
    "plt.plot(n_estimators_values[nmin:], score_rf_test[nmin:], label=\"rf test\")\n",
    "# plt.plot(n_estimators_values[nmin:], score_gb_train[nmin:], label=\"gb train\")\n",
    "plt.plot(n_estimators_values[nmin:], score_gb_test[nmin:], label=\"gb test\")\n",
    "plt.xlabel(\"n estimators\")\n",
    "plt.ylabel(\"$R^2$ score\")\n",
    "plt.legend()\n",
    "plt.title(\"GB better!!!\");"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Equally good"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "nmax = 2\n",
    "\n",
    "# plt.plot(n_estimators_values, score_rf_train, label=\"rf train\")\n",
    "plt.plot(n_estimators_values[:nmax], score_rf_test[:nmax], \"b\", label=\"rf test\")\n",
    "plt.ylabel(\"RF $R^2$ score\")\n",
    "plt.legend(loc=\"lower right\")\n",
    "plt.ylim(0.8, 0.9)\n",
    "plt.twinx()  # allows to have two y-axis with the same x-axis\n",
    "# plt.plot(n_estimators_values, score_gb_train, label=\"gb train\")\n",
    "plt.plot(n_estimators_values[:nmax], score_gb_test[:nmax], \"r\", label=\"gb test\")\n",
    "plt.xlabel(\"n estimators\")\n",
    "plt.ylabel(\"GB $R^2$ score\")\n",
    "plt.legend(loc=\"upper left\")\n",
    "plt.ylim(-0.01, 0.70)\n",
    "plt.title(\"Both equally good!!!\");"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "<br><br>\n",
    "\n",
    "**Be critical of your visualizations and try to make them as honest as possible.**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "<br><br>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## What did we learn today?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Principles of effective communication\n",
    "  - Concepts then labels, not the other way around.\n",
    "  - Bottom-up explanations.\n",
    "  - New ideas in small chunks.\n",
    "  - Reuse your running examples.\n",
    "  - Approaches from all angles.  \n",
    "  - When experimenting, show the results asap.\n",
    "  - **It's not about you.**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "- Decision variables, objectives, and context.\n",
    "- How does ML fit in?\n",
    "- Expressing your confidence about the results\n",
    "- Misleading visualizations.  "
   ]
  }
 ],
 "metadata": {
  "celltoolbar": "Slideshow",
  "kernelspec": {
   "display_name": "Python [conda env:cpsc330]",
   "language": "python",
   "name": "conda-env-cpsc330-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}