Financial Loan Risk Assessment

Overview

This group project was developed with the goal of assessing financial loan risks using a dataset sourced from Kaggle. This dataset includes fictional data, consisting of approximately 32,000 records and 12 columns, where each row represents one loan record for one client.

Data Source

Source: Kaggle - Credit Risk Dataset
Rows: 32,000
Columns: 12

Data Preparation

Cleaning

Removed duplicate rows.
Deleted age values greater than 100.
Removed employment length values over 80 years.
Dropped rows with missing values (NaNs).
Excluded records where the interest rate equals 0%.

Transformations

Performed one-hot and label encoding on categorical data.
Encoded cb_person_default_on_file with Y/N to 0/1.
Transformed loan_grade from A to E.
One-hot encoded person_home_ownership categories (Own, Rent, Mortgage).
Encoded various loan_intent.

Train/Test Split and Scaling

Split data into training and testing sets with a test size of 20% and a random state of 0.
Applied MinMaxScaler for normalization.

Models Tested

Classification Models

Targets:

Loan Status: Identified profiles of clients more or less likely to default.
cb_person_default_on_file: Profiled historically defaulting clients.

Regression Models

Target:

Loan Amount: Explored the association of loan amount with other numerical indicators.

Model Selection

Employed a weighted average F1 Score for classification models.
Used weighted average R2 Score for regression models.
Selected models include KNN, Decision Tree, Logistic Regression, Bagging & Pasting, Random Forest, Ada Boosting, and Gradient Boosting.

Key Findings

Application: Loan Status Target

Low Default Risk Profile

Higher income, lack of previous defaults, home mortgage, and homeownership.

High Default Risk Profile

Higher loan percentage of income, renters, previous defaults, and higher interest rates.

Application: Loan Amount

Developed targeted loan products and marketing strategies based on customer segmentation by age, income, and loan reason.
Formulated strategies for interest rate assessments based on loan amounts.

Conclusion

The project successfully implemented multiple machine learning models to assess and predict financial loan risks. The Random Forest model showed the highest accuracy in predicting loan amounts, while the Random Patches and Gradient Boosting models excelled in the loan status classification.

darceeanne / financial_loan_risk_assessment Goto Github PK