Classification problem for Loan Default Prediction

This project aims to predict loan approval statuses for for a Non-Banking Financial Company (NBFC) based on transaction and customer data using machine learning techniques. The goal is to build a model that accurately classifies whether a loan application will be approved or denied, based on historical transactional data

Task Summary:

The primary goal of NBFC is to:

• Develop a classification model to predict loan repayment behavior.

• Identify potential defaulters and non-defaulters.

• Enhance risk assessment and improve the loan approval process.

Repository Structure

DSW-Classification-Problem
├── Dataset
│   ├── test_data.xlsx
│   ├── train_data.xlsx
├── Problem Statement
│   ├── DSW_ML_Problem_Statement.pdf
├── Solution Notebooks
│   ├── eda.ipynb
│   ├── ModelTraining_Evaluation.ipynb
├── Rutuja Patil.zip
├── LICENSE
└── README.md

Data Overview:

• Historic data: Loan disbursement application and their default and non-default status for past 2 years+ has been kept in the file.

File name: train_data.xlsx

• Validation data: Loan disbursement application and their default and non-default status for past 3 months has been kept in the file.

File name: test_data.xlsx

Key Features:

customer_id: Represents the unique identifier for the customer.
transaction_date: The date of the transaction.
sub_grade: Sub-grade information.
term: Loan term details.

Target variable:
loan_status: (1 for default, 0 for non-default)

Workflow

1. Exploratory Data Analysis (EDA):

   • Analyze the dataset to uncover patterns and relationships.

   • Visualize key features and explore correlations with loan repayment behavior.

2. Data Preprocessing:

   • Clean and preprocess the data, including handling missing values and outliers.

   • Parse datetime features (e.g., transaction date) and extract relevant information like year, month, day, hour, and minute.

3. Feature Engineering:

   • Encode categorical variables (e.g., loan status, customer demographics) into numerical formats.

   • Split the data into input features (X) and target variable (y), where the target is the loan repayment behavior (default or non-default).

4. Model Selection & Training:

   • Classification models USED: Decision Tree, Logistic Regression, XGBoost, KNN and Naive Bayes

   • Used an object-oriented, class-based approach with methods for data loading, preprocessing, training, testing, and prediction.

   • Tune hyperparameters using GridSearchCV for optimal model configuration.

5. Model Evaluation:

   • Evaluate model performance on the test set using metrics like accuracy, confusion matrix, precision, recall, and F1-score.

   • Assess the model's ability to classify defaulters and non-defaulters accurately.

Installation and Usage

Ensure you have the following installed:

• Python 3.7+

• pip

Setup

To clone this repository, run the following command in your terminal:

git clone https://github.com/Rutuja1193/DSW-Classification-Problem.git

Navigate to the project directory:

cd DSW-Classification-Problem

Run the notebooks:

• EDA: Solution notebooks/eda.ipynb

• Model Training & Evaluation: Solution notebooks/ModelTraining_Evaluation.ipynb