The goal of this project is to predict whether a person’s credit score will increase, decrease, or stable based on their income, expenses, and personal details. This can help banks or financial companies better understand customer behavior and reduce financial risk.
We created a realistic synthetic dataset with 25,000 records. Each record represents a person with features such as Age, Location, Gender, Monthly Income, EMI payments, Credit Utilization Ratio, etc.
Values were generated with realistic distributions and location-based ranges gathered from publicly available sources. For example, people in metropolitan areas had higher average incomes and older age profiles.
| Feature | Metropolitan | Urban | Semi-Urban | Rural |
|---|---|---|---|---|
| Average Age | 28 | 30 | 29 | 26 |
| Age Range | 21–60 | 20–65 | 18–60 | 15–58 |
| Gender Ratio | 3:1 → 75% M | 5:2 → 71% M | 2:1 → 67% M | 5:4 → 56% M |
| Average Income | ₹1.5L | ₹90k–₹1L | ₹50k–₹75k | ₹20k–₹25k |
| Income Range | ₹15k–₹5L | ₹10k–₹4L | ₹8k–₹2.5L | ₹5k–₹2L |
- Removed outliers
- Performed label encoding for categorical features (e.g., Gender, Location)
- Used SMOTE (Synthetic Minority Oversampling Technique) on the training data to handle class imbalance
- Trained multiple models: Logistic Regression, Decision Tree, and Random Forest
- Decision Tree was chosen due to its superior performance on imbalanced data
- Used GridSearchCV for hyperparameter tuning
- Accuracy: 97.9%
- F1 Score: 98.07%
- The "Stable" class was the most frequent, and SMOTE + tuning helped ensure balanced performance across all classes
This model helps identify credit behavior patterns across various customer segments based on income, repayment history, and credit usage:
- Most people’s credit scores remain stable, as expected.
- Banks can proactively identify risky customers and improve offerings for low-risk and stable users.
1. High-Risk Customers:
- Offer flexible repayment plans
- Temporarily lower credit card limits
- Limit loan offerings until behavior improves
2. Stable Customers:
- Provide quick access to personal loans or credit cards
- Recommend safe financial products
- Reward consistent repayment behavior annually
3. Low-Risk Customers:
- Increase credit limits or reduce interest rates
- Encourage long-term savings through special plans
- Introduce "Buy Now, Pay Later" options
- Reward major credit milestones with vouchers
credit_data.csv– Full synthetic datasetLasya_credit_project.ipynb– Main project code notebook (modeling + insights)
Lasya