Financial-Risk-Score-Prediction

A Financial Risk Score is a numerical representation of a customer's potential risk to a lender. It helps banks, financial institutions, and credit agencies assess the likelihood that a customer will default on a loan or default on their financial obligations.

The following processes were followed in this project:

1) Data Collection:

Data was taken from Kaggle

2) Data Preprocessing:

• Determination of categorical and numerical features
• Feature engineering
• Digitization of categorical data (Ordinal Encoding and One-Hot Encoding)
• Normalization (Standard Scaler)

3) Model Selection and Training:

• XGBoost hyperparameter optimization (Just added parameters)
• LightGBM hyperparameter optimization (Just added parameters)
• CatBoost hyperparameter optimization (Just added parameters)
• Establishing an ensemble model (VotingRegressor)

4) Model Evaluation:

• Performance analysis with metrics such as "Mean Squared Error", "Root Mean Squared Error", "Mean Absolute Error", "R2 Score".

5) API:

• Creating an API with fastapi framework

Installation

Clone the repository:

git clone https://github.com/hakankocakk/Financial-Risk-Score-Prediction.git

Install dependencies:

pip install -r requirements.txt

Remove git tracking from src/model folder:

git rm -r --cached 'src/model'

Run the pipeline:

dvc repro

Show model performance

dvc metrics show

Testing the model with API

Start FastAPI application:

uvicorn src.api.main:app --reload

Open API documentation page: API Documentation

Sending a POST request: Click the POST button and then the Try it out button.

{
  "Age": 32,
  "AnnualIncome": 32097,
  "CreditScore": 586,
  "EmploymentStatus": "Employed",
  "EducationLevel": "Bachelor",
  "Experience": 0,
  "LoanAmount": 35206,
  "LoanDuration": 60,
  "MaritalStatus": "Single",
  "NumberOfDependents": 2,
  "HomeOwnershipStatus": "Rent",
  "MonthlyDebtPayments": 662,
  "CreditCardUtilizationRate": 0.475820793978558,
  "NumberOfOpenCreditLines": 4,
  "NumberOfCreditInquiries": 2,
  "DebtToIncomeRatio": 0.4553739117467006,
  "BankruptcyHistory": 0,
  "LoanPurpose": "Debt Consolidation",
  "PreviousLoanDefaults": 0,
  "PaymentHistory": 24,
  "LengthOfCreditHistory": 14,
  "SavingsAccountBalance": 611,
  "CheckingAccountBalance": 1091,
  "TotalAssets": 40962,
  "TotalLiabilities": 2852,
  "MonthlyIncome": 2674,
  "UtilityBillsPaymentHistory": 0.97994463185461,
  "JobTenure": 6,
  "NetWorth": 38110,
  "BaseInterestRate": 0.2472059999999999,
  "InterestRate": 0.2512794061689278,
  "MonthlyLoanPayment": 1035.9850545848378,
  "TotalDebtToIncomeRatio": 0.634820097050131
}

Add this text and click Execute button:
You can see the prediction result in the Response body section.

In the terminal where you are running the API application, stop the API by pressing CTRL + C.

About the Developer

• Hakan KOCAK