Banner source

Badge source

Key findings: People with the highest income, and who have at least one partner, are more likely to be approved for a credit card.

Authors

• @semasuka

Table of Contents

• People with the highest education level, and who are either husbands or wifes make more money
• Authors
• Table of Contents
• Business problem
• Data source
• Methods
• Tech Stack
• Quick glance at the results
• Lessons learned and recommendation
• Limitation and what can be improved
• Run Locally
• Explore the notebook
• Deployment on streamlit
• App deployed on Streamlit
• Repository structure
• Contribution
• License

Business problem

This app predicts if an applicant will be approved for a credit card or not. Each time there is a hard enquiry your credit score is affected negatively. This app predict the probability of being approved without affecting your credit score. This app can be used by applicant who wants to find out if they will be approved for a credit card without affecting their credit score.

Data source

• Kaggle credit card approval prediction

Methods

• Exploratory data analysis
• Bivariate analysis
• Multivarate correlation
• S3 bucket model hosting
• Model deployment

Tech Stack

• Python (refer to requirement.txt for the packages used in this project)
• Streamlit (interface for the model)
• AWS S3 (model storage)

Quick glance at the results

Correlation between the features.

Confusion matrix of gradrient boosting classifier.

ROC curve of gradrient boosting classifier.

Top 3 models (with default parameters)

Model	Recall score
Support vector machine	88%
Gradient boosting	90%
Adaboost	79%

• The final model used is: Gradient boosting

• Metrics used: Recall

• Why choose precision as metrics: Since the objective of this problem is to minimize the risk of credit default for the financial institution, the metrics to use depends on the current economical situation:

  • During the time of a bull market (when the economy is expending), people feel wealthy and usually are employed. Money is usually cheap and the risk of default is low. The financial institution is able to handle the risk of default therefore is not very strict on giving out credit. The financial institution can handle a number of bad clients as long as the vast majority of applicants are good clients (aka those who payback their credit).In this case, having a good recall (sensitivity) is ideal.

  • During a bear market (when the economy is contracting), people loose their jobs and their money through the stock market. Many people struggle to meet their financial obligations. The financial institution therefore tend to be more conservative on giving out credit or loans. The financial institution can't afford to give out credit to clients who won't be able to pay back their credit. The financial institution would rather have a smaller number of good clients even if it means that some good clients where denied credit, and ideally not have any bad client. In this case, having a good precision (specificity) is desirable.

    Note: There is always a trade-off between precision and recall. Choosing the right metrics depends on the problem you are solving.

    Conclusion: In our case, since we are in the longest bull market (not including the March 2020 flash crash), we will use recall as our metric.

Lessons learned and recommendation

• Based on the analysis on this project, we found out that the education level and type of relationship are the most predictive features to determine if someone makes more or less than 50K. Other features like Capital gain, hours work and age are also usefull. The least usefull features are: their occupation and the workclass they belong to.
• Recommendation would be to focus more on the most predictive feature when looking at the applicant profile, and pay less attention on their occupation and workclass.

Limitation and what can be improved

• Speed: since the model is stored on AWS S3, it can take some few seconds to load. Solution: cache the model with the Streamlit @st.experimental_singleton for faster reload.
• Dataset used: the dataset used is from 1990, inflation has not been taken into consideration and the countries's economies have changed since then. Solution: retrain with a more recent dataset.
• Hyperparameter tuning: I used RandomeSearchCV to save time but could be improved by couple of % with GridSearchCV.

Run Locally

Initialize git

git init

Clone the project

git clone https://github.com/semasuka/Credit-card-approval-prediction-classification.git

enter the project directory

cd Credit-card-approval-prediction-classification

Create a conda virtual environment and install all the packages from the environment.yml (recommended)

conda env create --prefix <env_name> --file assets/environment.yml

Activate the conda environment

conda activate <env_name>

List all the packages installed

conda list

Start the streamlit server locally

streamlit run cc_approval_pred.py

If you are having issue with streamlit, please follow this tutorial on how to set up streamlit

Explore the notebook

To explore the notebook file here

Deployment on streamlit

To deploy this project on streamlit share, follow these steps:

• first, make sure you upload your files on Github, including a requirements.txt file
• go to streamlit share
• login with Github, Google, etc.
• click on new app button
• select the Github repo name, branch, python file with the streamlit codes
• click advanced settings, select python version 3.9 and add the secret keys if your model is stored on AWS or GCP bucket
• then save and deploy!

App deployed on Streamlit

Repository structure

├── datasets
│   ├── GDP.csv                     <- the data used to feature engineering/enriched the original data.
│   ├── test.csv                    <- the test data.
│   ├── train.csv                   <- the train data.
│
│
├── assets
│   ├── confusion_matrix.png        <- confusion matrix image used in the README.
│   ├── gif_streamlit.gif           <- gif file used in the README.
│   ├── heatmap.png                 <- heatmap image used in the README.
│   ├── Income_classification.png   <- banner image used in the README.
│   ├── environment.yml             <- list of all the dependencies with their versions(for conda environment).
│   ├── roc.png                     <- ROC image used in the README.
│
├── pandas_profile_file
│   ├── income_class_profile.html   <- exported panda profile html file.
│
│
├── .gitignore                      <- used to ignore certain folder and files that won't be commit to git.
│
│
├── Income_Classification.ipynb     <- main python notebook where all the analysis and modeling are done.
│
│
├── LICENSE                         <- license file.
│
│
├── income_class_st.py              <- file with the best model and best hyperparameter with streamlit component for rendering the interface.
│
│
├── README.md                       <- this readme file.
│
│
├── requirements.txt                <- list of all the dependencies with their versions(used for Streamlit ).

Contribution

Pull requests are welcome! For major changes, please open an issue first to discuss what you would like to change or contribute.

License

MIT License

Copyright (c) 2022 Stern Semasuka

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Learn more about MIT license