Skip to content

Latest commit

 

History

3 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 

Repository files navigation

🏦 Enhancing Loan Approval Accuracy through Ensemble-Based Machine Learning

Python Conference Publisher Models Best Accuracy

Research paper presented at the 4th Congress on Smart Computing Technologies (CSCT 2025) hosted by National Institute of Technology, Sikkim, India — December 13–14, 2025. Published under Springer.

Authors: Mahamat Hanga Derio, Chaitra P C, Sujatha Arun Kokatnoor Christ University, Bangalore, India


📌 Research Objective

Loan approval decisions in financial institutions are often inconsistent, biased, or slow when handled manually. This study systematically benchmarks 10 ensemble machine learning algorithms to identify the most accurate and reliable model for automated loan approval prediction — with implications for microfinance institutions in data-scarce economies.


📊 Dataset

  • Source: Kaggle — Loan Status Prediction
  • Features: Gender, Marital Status, Dependents, Education, Employment, Income, Co-applicant Income, Loan Amount, Loan Term, Credit History, Property Area
  • Target: Loan Status (Approved / Rejected)
  • Preprocessing: Mode imputation (categorical), median imputation (numerical), IQR-based outlier removal, Label Encoding

🤖 Models Benchmarked (10 Algorithms)

Model Accuracy
LightGBM 89% ✅ Best
Random Forest 88%
CatBoost 87%
XGBoost 86%
Extra Trees 86%
AdaBoost 85%
Gradient Boosting 84%
Bagging Classifier 84%
Stacking Classifier 82%
Voting Classifier 79%

Key finding: Gradient boosting variants (LightGBM, CatBoost, XGBoost) consistently outperform bagging-based and meta-ensemble approaches on this tabular financial dataset. LightGBM's leaf-wise tree growth gives it an edge in handling the income skew and categorical feature interactions present in loan data.


🔬 Methodology

Raw Loan Dataset (Kaggle)
    ↓
Preprocessing — Imputation + Outlier Removal + Label Encoding
    ↓
Train/Test Split (80/20, random_state=42)
    ↓
Feature Importance Analysis (Random Forest)
    ↓
10 Ensemble Models Trained & Evaluated
    ↓
Accuracy + Confusion Matrix + Classification Report per model
    ↓
Final Comparison — Bar chart visualization

📌 Key Features Used

Credit History emerged as the strongest predictor of loan approval, followed by Applicant Income and Loan Amount — consistent with lending theory and prior literature.


🚀 How to Run

# Clone
git clone https://github.com/Derio001/loan-prediction-ensemble-models.git
cd loan-prediction-ensemble-models

# Install dependencies
pip install pandas numpy matplotlib seaborn scikit-learn xgboost lightgbm catboost kagglehub

# Run notebook
jupyter notebook The_Project_Implementation.ipynb

Dataset is auto-downloaded via kagglehub — no manual download needed.


🏛️ Conference & Publication

Detail Info
Conference 4th Congress on Smart Computing Technologies (CSCT 2025)
Venue National Institute of Technology, Sikkim, India
Dates December 13–14, 2025
Paper ID 872
Publisher Springer
Certificate Presented by Mahamat Hanga Derio

👤 Authors

Mahamat Hanga Derio — M.Tech Data Science, Christ University, Bangalore Chaitra P C — Christ University, Bangalore Dr. Sujatha Arun Kokatnoor — Department of AIML & DS, Christ University, Bangalore

🔗 GitHub | LRI Health Project | GDP Intelligence Project


Part of ongoing research in applied machine learning for financial and development contexts.

About

Improving loan approval prediction using ensemble machine learning models with robust preprocessing and imbalance handling.

Resources

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages