🌟 Loan Approval Prediction Project 🌟

📖 Overview

This project aims to develop a predictive model that determines the likelihood of loan approval based on various features of loan applicants. By leveraging machine learning techniques, we analyze historical loan data to predict outcomes for new loan applications.

📁 Project Structure

Loan-approval-prediction/
│
├── 📓 LoanApprovalEDA.ipynb          # Jupyter Notebook for Exploratory Data Analysis
├── 📦 final_model.pkl                # Trained model for loan approval predictions
├── 📦 loan_approval_model.pkl         # Model used for predictions
├── 📊 loan_approval_predictions.csv    # Predictions made on new data
├── 🧪 test_Y3wMUE5_7gLdaTN.csv       # Test dataset
└── 📚 train_u6lujuX_CVtuZ9i.csv      # Training dataset

📊 Data Description

Training Data

The training dataset consists of 614 entries and the following columns:

Column	Description
Loan_ID	Unique identifier for each loan
Gender	Gender of the applicant
Married	Marital status of the applicant
Dependents	Number of dependents
Education	Educational qualification
Self_Employed	Employment status of the applicant
ApplicantIncome	Income of the applicant
CoapplicantIncome	Income of the coapplicant
LoanAmount	Amount of loan applied for
Loan_Amount_Term	Duration of the loan in months
Credit_History	Credit history of the applicant
Property_Area	Area of residence
Loan_Status	Approval status (Y/N)

Test Data

The test dataset has the same structure as the training dataset but does not contain the Loan_Status column.

📈 Exploratory Data Analysis (EDA)

• Conducted statistical analysis to understand the distribution of features.
• Visualized the loan approval status distribution, highlighting class imbalance.
• Identified categorical and numerical columns for further processing.

🔍 Data Preprocessing

1. Missing Value Imputation:

   • Imputed missing values in LoanAmount, Loan_Amount_Term, and Credit_History using appropriate strategies.

2. Encoding Categorical Variables:

   • Converted categorical features into numerical representations using one-hot encoding.

3. Scaling:

   • Scaled numerical features to normalize their distributions.

4. Feature Selection:

   • Dropped unnecessary columns such as Loan_ID before model training.

🏗️ Model Building

1. Train-Test Split:

   • Split the training data into training and validation sets.

2. Model Selection:

   • Evaluated various models, including Logistic Regression, Decision Tree, and Random Forest.

3. Hyperparameter Tuning:

   • Utilized GridSearchCV for hyperparameter tuning of the Random Forest model.

4. Model Evaluation:

   • Evaluated models based on accuracy, precision, recall, and F1-score.

📊 Results

Model	Accuracy	Key Parameters
Logistic Regression	79%	-
Decision Tree	68%	-
Random Forest	78%	max_depth: None, max_features: 'sqrt', min_samples_leaf: 1, min_samples_split: 10, n_estimators: 50

Final Model Accuracy: 78.86%

🚀 Future Work

• Address Class Imbalance: Implement techniques to address class imbalance to improve model performance.
• Feature Engineering: Explore additional features that could enhance prediction accuracy.
• Deployment: Create a web application for real-time predictions of loan approvals.

🛠️ Usage

To use the trained model for making predictions, load the final_model.pkl and preprocess your input data similarly to the training data preprocessing steps.

import pandas as pd
import joblib

# Load model
model = joblib.load('final_model.pkl')

# Preprocess new data
new_data_processed = preprocess(new_data)

# Make predictions
predictions = model.predict(new_data_processed)

📝 Acknowledgments

• Pandas for data manipulation
• Scikit-learn for machine learning algorithms
• Matplotlib and Seaborn for data visualization