ML-DS-Playground

This repository is a collection of personal machine learning and data science projects created while learning and experimenting with various techniques. The projects included here cover a wide range of topics such as supervised and unsupervised learning, clustering, feature engineering, and more.

Projects

Here is a list of some projects included in this repository:

1. Supervised Learning

   • Classify Song Genres from Audio Data
     • Techniques: Exploratory Data Visualization, Feature Scaling, Feature Reduction, Cross-Validation, Logistic Regression, Decision Trees
     • Tasks: Classification of song genres based on audio features, model evaluation using classification reports.

2. Unsupervised Learning

   • Clustering Antarctic Penguin Species:
     • Techniques: Data Manipulation with Pandas, Dummy Variable Creation, Elbow Method, K-Means Clustering, Feature Scaling
     • Tasks: Clustering penguin species based on features like flipper length and body mass.

3. Feature Engineering

   • Exploring NYC Public School Test Result Scores
     • Techniques: Sorting and Subsetting, Grouped Summary Statistics
     • Tasks: Feature engineering to understand NYC public school test results.

4. Deep Learning with Keras

   • Building an E-Commerce Clothing Classifier Model with Keras
     • Techniques: Neural Networks in Keras, Compiling and Training Models, CNNs, Image Classification
     • Tasks: Classification of clothing items in e-commerce datasets using convolutional neural networks.

5. Natural Language Processing and Audio Analysis Techniques

   • Customer Support Calls
     • Techniques: Speech Recognition, Audio Feature Extraction, Sentiment Analysis, Named Entity Recognition (NER), Text Similarity/Embedding
     • Tasks: Analyzing customer support calls to identify sentiment, entities, and key topics from audio.

6. Data Engineering for E-commerce Orders and Demand Forecasting with PySpark

   • Cleaning an Orders Dataset with PySpark

     • Techniques: Data Cleaning, PySpark, Data Transformation
     • Tasks:
       • Cleaning and preprocessing an e-commerce orders dataset.
       • Removed orders placed between 12 am and 5 am.
       • Created a time_of_day column.
       • Filtered out products that are no longer sold.
       • Converted columns to lowercase and extracted relevant address data.
       • Exported cleaned data for use in demand forecasting.

   • Building a Demand Forecasting Model

     • Techniques: Random Forest Regression, Feature Engineering with PySpark, Time-Series Forecasting
     • Tasks:
       • Forecasting sales and inventory needs for promotional planning.
       • Cleaned and aggregated sales data at daily intervals.
       • Built and trained a Random Forest model to predict future product sales.
       • Evaluated model performance using Mean Absolute Error (MAE).
       • Predicted sales for promotional week and ensured optimal inventory management.

7. Recommendation Systems and Data Visualization

   • Comparing Cosmetics by Ingredients
     • Techniques: Content-Based Filtering, Ingredient Tokenization, Document-Term Matrix (DTM), Dimensionality Reduction with t-SNE, Interactive Visualization with Bokeh
     • Tasks: Built a recommendation system for moisturizers targeting dry skin by analyzing ingredient similarities. Used t-SNE for dimensionality reduction and Bokeh for interactive product comparison visualization.

Technologies Used

The projects in this repository are implemented using the following technologies:

• Programming Language: Python
• Interactive Environment: Jupyter Notebook
• Libraries for Data Manipulation and Analysis: Pandas, NumPy
• Machine Learning Frameworks: Scikit-Learn, TensorFlow/Keras (for deep learning projects), PySpark (for large-scale data processing and engineering)
• Visualization Tools: Matplotlib, Seaborn, Bokeh
• NLP Libraries: NLTK, SpaCy, SpeechRecognition (for speech-to-text)
• Clustering and Feature Engineering Techniques: K-Means Clustering, t-SNE, Document-Term Matrix (DTM)
• Specialized Techniques: Random Forest Regression for forecasting, Content-Based Filtering for recommendation systems