riziuzi/Loan-Predictor

Loan Prediction Project

Table of Contents

• Overview
• Dataset
• Data Preprocessing
• Exploratory Data Analysis
• Feature Engineering
• Model Development
• Results
• Installation
• Usage
• Contributing
• License

Overview

This project implements a machine learning pipeline for predicting loan approvals. It uses various techniques in data preprocessing, exploratory data analysis, feature engineering, and machine learning modeling to achieve high predictive accuracy.

Dataset

The project uses two datasets:

• loan_train.csv: Training dataset with 491 entries and 13 features
• loan_test.csv: Test dataset with 123 entries and 12 features (excluding the target variable)

Key features include:

• Loan_ID
• Gender
• Married
• Dependents
• Education
• Self_Employed
• ApplicantIncome
• CoapplicantIncome
• LoanAmount
• Loan_Amount_Term
• Credit_History
• Property_Area
• Loan_Status (target variable)

Data Preprocessing

1. Handling missing values:

   • Mode imputation for categorical variables (Gender, Married, Dependents, Self_Employed, Credit_History)
   • Median imputation for numerical variables (LoanAmount)
   • Mode imputation for Loan_Amount_Term

2. Outlier treatment:

   • Log transformation applied to LoanAmount to handle right skewness

3. Encoding categorical variables:

   • One-hot encoding used for all categorical variables

Exploratory Data Analysis

1. Univariate analysis:

   • Distribution of loan approval status
   • Distribution of categorical variables (Gender, Married, Dependents, Education, Self_Employed, Credit_History, Property_Area)
   • Distribution of numerical variables (ApplicantIncome, CoapplicantIncome, LoanAmount)

2. Bivariate analysis:

   • Relationship between categorical variables and loan approval status
   • Relationship between numerical variables and loan approval status

3. Correlation analysis:

   • Heatmap of correlation between numerical variables

Feature Engineering

Based on domain knowledge and insights from EDA, we created the following new features:

1. Total_Income: Combination of ApplicantIncome and CoapplicantIncome
2. Total_Income_log: Log transformation of Total_Income
3. EMI: LoanAmount divided by Loan_Amount_Term
4. Balance Income: Total_Income minus (EMI * 1000)

Model Development

Four models were implemented and evaluated:

1. Logistic Regression
2. Decision Tree
3. Random Forest (Pending)
4. XGBoost (Pending)

Each model was evaluated using 5-fold stratified cross-validation to ensure robust performance estimation.

Results

Here's a summary of the model performances:

Model	Mean Validation Accuracy	Mean Validation F1 Score	AUC
Logistic Regression	0.7881	0.8622	0.8176
Decision Tree	0.7168	0.8027	0.7500
Random Forest	Not implemented	Not implemented	N/A
XGBoost	Not implemented	Not implemented	N/A

Note: The Random Forest and XGBoost models were mentioned in the project plan but not implemented in the provided notebook.

Installation

git clone https://github.com/riziuzi/loan-prediction.git
cd loan-prediction
pip install -r requirements.txt

Contributing

Contributions to this project are welcome. Please fork the repository and submit a pull request with your proposed changes.

---

Created by riziuzi