Credit Score Prediction

This projet is a Midterm project of ML Zoomcamp by Alexey Grigorev

Overview

The extensive dataset of customer banking and credit-related details collected over the years. The goal is to develop a Classification ML model capable of categorizing individuals into predefined credit categories, thereby minimizing manual effort and improving efficiency. This project provides a machine learning-based app for predicting a person's credit score based on their financial and credit-related information. It uses a Flask web application to serve the trained model and Docker for containerization. The goal is to help a global finance company automate the classification of individuals into predefined credit score brackets, reducing manual effort and increasing efficiency.

Dataset contains 28 columns and 100K rows: Target feature is "credit_score" : "Good," "Poor," "Standard".

No	Feature	Description
1	id	Unique identifier for each record.
2	customer_id	Unique identifier for each customer.
3	month	Month of the transaction or record.
4	name	Customer’s name.
5	age	The customer’s age.
6	ssn	Customer’s social security number.
7	occupation	The customer’s occupation.
8	annual_income	The customer’s annual income.
9	monthly_inhand_salary	The customer’s monthly take-home salary.
10	num_bank_accounts	Total number of bank accounts owned by the customer.
11	num_credit_card	Total number of credit cards held by the customer.
12	interest_rate	The interest rate applied to loans or credits.
13	num_of_loan	Number of loans the customer has taken.
14	type_of_loan	Categories of loans obtained by the customer.
15	delay_from_due_date	The delay in payment relative to the due date.
16	num_of_delayed_payment	Total instances of late payments made by the customer.
17	changed_credit_limit	Adjustments made to the customer’s credit limit.
18	num_credit_inquiries	Number of inquiries made regarding the customer's credit.
19	credit_mix	The variety of credit types the customer uses (e.g., loans, credit cards).
20	outstanding_debt	Total amount of debt the customer currently owes.
21	credit_utilization_ratio	Proportion of credit used compared to the total credit limit.
22	credit_history_age	Duration of the customer’s credit history.
23	payment_of_min_amount	Indicates if the customer pays the minimum required amount each month.
24	total_emi_per_month	Total Equated Monthly Installment (EMI) paid by the customer.
25	amount_invested_monthly	Monthly investment amount made by the customer.
26	payment_behaviour	Customer’s payment habits and tendencies.
27	monthly_balance	The remaining balance in the customer’s account at the end of each month.
28	credit_score	The customer’s credit score (target variable: "Good," "Poor," "Standard").

Download the latest version

The dataset from Kaggle:

path = kagglehub.dataset_download("parisrohan/credit-score-classification")

Problem Statement

Managing credit risk is a critical task for financial institutions. Over the years, this company has gathered vast amounts of customer data, including banking and credit information. The task is to create an intelligent system capable of:

• Classifying individuals into credit score classes.
• Reducing the manual workload for the company's analysts.
• Delivering accurate predictions for better risk management.

Files description

The dataset cleaning, analysis and model selection were conducted using Jupyter Notebook. You can find the corresponding notebook file in this repository:

• notebook_cleaning and preprocessing.ipynb
• notebook_EDA_model_selection.ipynb

Cleaned dataset and dataset for testing final model:

• cleaned_df.csv
• df_test.csv

Trained final model with best parameters and pickled files:

• train.py
• final_model.pkl
• dict_vectorizer.pkl
• label_encoder.pkl

Files for running the project locally:

• predict.py
• predict-test.py
• requirements.txt

Running using Waitress as WSGI server

Please use predict.py and predict-test.py for testing model prediction.

1. Create and activate .venv:

  python -m venv venv
venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

4. Run predict.py:

waitress-serve --listen=0.0.0.0:9696 predict:app

5. Run predict-test.py and use the command line:

python predict-test.py