Loan Default Prediction Machine Learning Project

This is an exploratory project for me to apply and compare different ML models and techniques, including:

• Feature Engineering
  • One-hot encoding for categorical features
  • Normalization/standardization
  • Imputation
  • Feature expansion
  • Feature reduction:
    • Feature hashing
    • Feature selection
    • Principal component analysis
  • Feature discretization with Decision Trees or Random Forests
• Machine Learning Models:
  • Logistic Regression
  • Decision Trees, Random Forest, Gradient-Boosted Decision Trees
  • K-Nearest-Neighbor
  • Support Vector Machines
  • Neural Networks

The data is from a Kaggle competition Loan Default Prediction.

Dependencies

Python 3, numpy, pandas, scikit-learn, matplotlib, xgboost, tensorflow, keras.

Usage

Download train_v2.csv from https://www.kaggle.com/c/loan-default-prediction/data and put in the loan-default-prediction directory. Running the first Jupyter notebook, LDP 01 - Data Preprocessing.ipynb will give you a few processed CSV files, which are used in subsequent notebooks as the training data. The rest of the notebooks do not have strong dependencies.