Loan Approval Prediction uses machine learning to analyze factors like income and credit history, automating and improving the loan approval process. The project ensures accurate, efficient, and user-friendly predictions with a Gradio interface.
Loan approval prediction is a critical task in the banking and finance industry. This project leverages machine learning to predict whether a loan application will be approved based on features such as applicant income, credit history, loan amount, and more. By automating the decision-making process, this system aims to reduce risks for financial institutions and improve customer satisfaction.
- Machine Learning Models: Uses multiple algorithms like Random Forest, Logistic Regression, K-Nearest Neighbors, and Neural Networks.
- Dynamic Output: Displays prediction results with color-coded status (e.g., "Approved" in green, "Rejected" in red).
- User-Friendly Interface: Built with Gradio for real-time loan status prediction.
- Feature Analysis: Explores the impact of different features like credit history and income on loan approval.
- Programming Language: Python
- Libraries:
pandas,numpy: Data manipulation and preprocessingscikit-learn: Machine learning model developmentgradio: Building interactive interfacesmatplotlib,seaborn: Data visualization
- Description: The dataset includes key features such as:
- Applicant's Income
- Coapplicant's Income
- Loan Amount
- Credit History
- Property Area
- Source: (Mention if it’s from Kaggle, UCI, or custom data)
- Data Preprocessing:
- Handling missing values
- Encoding categorical variables
- Balancing classes with SMOTE
- Data Preprocessing: Cleaning, scaling, and encoding data for optimal model performance.
- Model Training:
- Trains multiple algorithms (e.g., Logistic Regression, Random Forest) and evaluates them using metrics like accuracy, precision, recall, and ROC-AUC.
- Deployment:
- Uses Gradio for an interactive prediction interface.
- Dynamic Output:
- Displays predictions with color-coded text to indicate loan status.
- Accuracy: Overall correctness of predictions.
- Precision: Reducing false positives (risky loans predicted as safe).
- Recall: Reducing false negatives (safe loans predicted as risky).
- ROC-AUC: Assessing model discrimination between approved and rejected loans.
The project evaluates multiple machine learning algorithms, including an Artificial Neural Network (ANN), to identify the best-performing model for loan approval prediction. Below are the performance metrics for each algorithm:
| Model Name | Accuracy (%) | Precision (%) | Confusion Matrix |
|---|---|---|---|
| Logistic Regression | 74.56 | 77.03 | [[52, 34], [9, 74]] |
| K-Nearest Neighbors | 79.29 | 74.94 | [[56, 30], [14, 69]] |
| Support Vector Machine | 74.56 | 77.47 | [[51, 35], [8, 75]] |
| Random Forest | 82.84 | 85.53 | [[60, 26], [3, 80]] |
| Decision Tree | 79.88 | 79.09 | [[60, 26], [11, 72]] |
| Artificial Neural Network | 77.78 | N/A | N/A |
- Training Process:
- Trained over 10 epochs using the Adam optimizer and binary cross-entropy loss function.
- Showed gradual improvement in accuracy and reduction in loss during training and validation.
- Performance Metrics:
- Training Accuracy: 77.06%
- Validation Accuracy: 77.78%
- Test Accuracy: 75.15%
- Random Forest achieves the highest accuracy (82.84%) and precision (85.53%), making it the most reliable model for this dataset.
- K-Nearest Neighbors and Decision Tree also perform well, with accuracy scores of 79.29% and 79.88%, respectively.
- Artificial Neural Network demonstrates competitive performance, showing the potential for further optimization with additional hyperparameter tuning or dataset expansion.
- Python 3.7+
- Install required libraries:
pip install -r requirements.txt
Contributions are welcome! Feel free to open issues or submit pull requests.
This project is licensed under the MIT License.