A machine learning project that applies Decision Tree and Random Forest algorithms to predict whether a borrower is likely to repay a loan. The project uses real lending-style data and compares tree-based models for classification performance.

This project analyzes borrower financial data and builds classification models using:

• Decision Tree Classifier
• Random Forest Classifier

The goal is to predict loan repayment behavior and compare how single-tree vs ensemble methods perform on the same dataset.

Predict whether a loan will be:

• Repaid ✅
• Default ❌

Using borrower and loan-related features.

• Python
• Pandas
• NumPy
• Scikit-learn
• Matplotlib
• Seaborn
• Jupyter Notebook

Based on publicly available lending-style data (LendingClub-type dataset).

Typical features include:

• Credit policy
• Interest rate
• Installment amount
• Annual income
• Debt-to-income ratio
• FICO score range
• Credit history metrics

Target variable:

loan_status / not_fully_paid

• Dataset loading and inspection
• Feature distribution analysis
• Class balance check
• Visualization of key variables

• Handling categorical variables
• Feature selection
• Train/Test split
• Encoding where required

• Train Decision Tree classifier
• Fit on training data
• Generate predictions
• Evaluate performance

• Train ensemble Random Forest classifier
• Compare against Decision Tree
• Analyze performance improvements

Models evaluated using:

• Accuracy
• Confusion Matrix
• Precision / Recall
• F1 Score
• Classification Report

For Decision Tree classification_report:

              precision    recall  f1-score   support

           0       0.84      0.83      0.84      2404
           1       0.19      0.21      0.20       470

    accuracy                           0.73      2874
   macro avg       0.52      0.52      0.52      2874
weighted avg       0.74      0.73      0.73      2874

For Random Forest classification_report:

              precision    recall  f1-score   support

           0       0.84      0.99      0.91      2404
           1       0.52      0.03      0.06       470

    accuracy                           0.84      2874
   macro avg       0.68      0.51      0.49      2874
weighted avg       0.79      0.84      0.77      2874

For the same dataset, Random Forest showed improved generalization and lower variance.

git clone https://github.com/rohitb281/decision-trees-random-forest-project.git
cd decision-trees-random-forest-project

pip install pandas numpy scikit-learn matplotlib seaborn

jupyter notebook

Decision Trees and Random Forest Project.ipynb

• Run all cells.

• Tree-based machine learning models
• Ensemble learning
• Bias–variance tradeoff
• Feature-based classification
• Model comparison methodology
• Evaluation metrics for classifiers

• Hyperparameter tuning (GridSearchCV)
• Cross-validation
• Feature importance analysis
• ROC–AUC curves
• Class imbalance handling
• Model explainability (SHAP)

• Limited hyperparameter tuning
• No production deployment
• Interpretability vs performance tradeoff

Open for educational and portfolio use.

• Rohit Bollapragada
• GitHub: https://github.com/rohitb281