Credit Risk Classification – Predictive Modeling & Evaluation

Project Overview

This project analyzes a credit risk dataset to understand the relationships between customer attributes and credit outcomes, and to build classification models capable of predicting a chosen binary credit risk variable.

The analysis is structured in two main stages:

1. Exploratory Data Analysis (EDA) to examine variable distributions, relationships, and data quality
2. Supervised classification modeling to assess predictive performance using multiple machine learning algorithms

---

Objectives

• Perform exploratory data analysis to understand feature distributions and relationships
• Handle missing values and mixed data types
• Build and compare multiple classification models:
  • Logistic Regression
  • Random Forest
  • Support Vector Machine (SVM)
  • Linear SVM
• Evaluate model performance using appropriate classification metrics
• Identify the most effective model for credit risk prediction

---

Dataset

Data Source

• credit_risk_dataset_classification.csv

The dataset contains customer-level financial and demographic features along with a binary target variable representing credit risk.

---

Technologies & Libraries Used

This project is implemented in Python using the following libraries:

Core Data Handling

• numpy
• pandas

Visualization

• matplotlib
• seaborn

Preprocessing & Pipelines

• scikit-learn
  • ColumnTransformer
  • Pipeline
  • OneHotEncoder
  • StandardScaler
  • SimpleImputer

Modeling

• LogisticRegression
• RandomForestClassifier
• SVC
• LinearSVC
• DecisionTreeClassifier

Model Selection & Evaluation

• train_test_split
• StratifiedKFold
• GridSearchCV
• ROC-AUC
• Precision-Recall AUC
• F1-score

Statistical Utilities

• scipy.stats (loguniform)
• matplotlib.ticker