Data Science Projects

This repository collects notebook-based machine learning and deep learning experiments across tabular data, image classification, natural language processing, recommendation systems, and time series forecasting.

Project Map

Area	Contents
Classification/	Credit decisions, customer segmentation labels, game win-rate prediction, AutoML classification, and survey-style classification notebooks.
Clustering/	Customer clustering and a small clustering walkthrough.
Computer Vision/	CNN and TensorFlow examples for CIFAR-10, Fashion MNIST, Kannada MNIST, digit recognition, parasite images, and traffic signs.
Natural Language Processing (NLP)/	Sentiment, review, disaster tweet, duplicate-question, and Amazon review classification notebooks.
Recommendation System/	Cosine-similarity examples for Netflix and TEDx-style recommendations plus a product recommendation notebook.
Regression/	House pricing and advertising regression examples.
Time Series/	Stock and ridership forecasting notebooks.
zingat house pricing/	Web scraping and regression notebooks for housing-price analysis.

The root defect-prediction-xgboost-83.ipynb notebook is kept at the top level because it is a standalone XGBoost defect prediction experiment.

Working With The Notebooks

Most projects are self-contained Jupyter notebooks. A typical local workflow is:

python -m venv .venv
.venv\Scripts\activate
python -m pip install jupyter pandas numpy scikit-learn matplotlib seaborn
jupyter notebook

Some notebooks use additional libraries such as TensorFlow, XGBoost, PySpark, or AutoML packages. Install those only for the notebook you plan to run so the base environment stays small.

Maintenance Notes

• Keep new notebooks inside the closest topic folder instead of adding more top-level files.
• Add a short markdown introduction near the top of each notebook that names the dataset, target variable, and evaluation metric.
• Avoid committing generated datasets, model artifacts, or local notebook checkpoint folders.
• Before committing notebook changes, restart the kernel and run all cells so outputs match the current code.