Loan Prediction Using Logistic Regression A data science project to predict whether a customer will accept a personal loan based on demographic and financial information.

Dataset: The dataset includes features such as:

Age, Experience, Income, Family, Education Credit card spending (CCAvg), Mortgage amount Online banking, CreditCard, CD/Investment account indicators Source: (https://www.kaggle.com/datasets/itsmesunil/bank-loan-modelling)

Exploratory Data Analysis (EDA) Investigated missing values, distributions, and feature correlations. Visualized loan approval rates by income, education, and account types. Model Performance Used LogisticRegression with preprocessing:

Pipeline: SimpleImputer → StandardScaler → LogisticRegression max_iter: 1000 (to ensure convergence) 🔢 Results: Metric Class 0 (No Loan) Class 1 (Loan Accepted) Precision 0.96 0.85 Recall 0.99 0.66 F1-Score 0.97 0.74 Overall Accuracy: 95% Macro Avg F1-Score: 0.86 Weighted Avg F1-Score: 0.95 ✅ This model balances accuracy and recall well, despite mild class imbalance.

Feature Importance (SHAP) Used SHAP to identify top features influencing model predictions:

SHAP Importance

Income and Education are the most influential. Online, Age, and Experience have minor contributions. Tech Stack Python (Pandas, NumPy, scikit-learn, SHAP, Matplotlib) Jupyter Notebook GitHub for version control and portfolio showcase Future Improvements Apply SMOTE or class-weight adjustments for better recall. Try ensemble methods like XGBoost. Deploy using Streamlit.