Bankruptcy Prediction with Decision Trees

📄 Project Overview

This project implements a machine learning solution to predict corporate bankruptcy using financial data. The core of this project is a Decision Tree classifier, a powerful and interpretable model, trained to distinguish between healthy and bankrupt companies.

The goal is to build a reliable predictive model that can serve as an early warning system for financial distress, demonstrating a practical application of machine learning in risk management and financial analysis. The entire workflow, from data exploration to model evaluation, is documented in the accompanying Jupyter Notebook.

✨ Key Features

• Exploratory Data Analysis (EDA): A thorough investigation of the dataset to understand feature distributions and relationships between financial metrics and the target variable (bankruptcy).
• Data Preprocessing: Preparing the data for modeling, including handling missing values and splitting the data into training and testing sets.
• Model Training: Building and training a Decision Tree Classifier using Scikit-learn.
• Hyperparameter Tuning: Optimizing the Decision Tree model's parameters to improve performance and prevent overfitting.
• Performance Evaluation: Assessing the model's accuracy, precision, recall, and F1-score, and visualizing the results with a confusion matrix and classification report.

📊 Dataset

The project utilizes the Bankrupcy_data.csv dataset. It contains various financial ratios and metrics from a collection of companies, with a binary target variable indicating whether a company went bankrupt or not.

• Target Variable: Bankrupt? (1 for bankrupt, 0 for healthy)
• Features: Includes metrics such as return on equity, debt ratio, and other financial indicators.

💻 Technologies Used

• Language: Python 3
• Libraries:
  • Pandas & NumPy (for data handling and numerical operations)
  • Matplotlib & Seaborn (for data visualization)
  • Scikit-learn (for machine learning model implementation and evaluation)
  • Jupyter Notebook (as the development environment)

🚀 Getting Started

Prerequisites

• Python 3.8 or newer
• Jupyter Notebook or JupyterLab
• Git

Installation and Usage

1. Clone the repository:

   git clone https://github.com/Kamaranis/Classification-with-decision-trees.git
   cd Classification-with-decision-trees

2. Install the necessary libraries: For best practice, create a virtual environment first.

   pip install pandas numpy scikit-learn matplotlib seaborn jupyter

3. Launch Jupyter Notebook:

   jupyter notebook

4. Run the analysis: Open the Decision Tree.ipynb file and execute the cells in order to see the full analysis, model training process, and results.

📈 Results Showcase

Figure 1: Credit type and number of non-payments

Figure 2: Default rate by credit purpose

Final Model Performance:

Confusion Matrix:

Figure 3: Confusion matrix I

Figure 4: Confusion matrix II _

👤 Author

Antonio Barrera Mora

• LinkedIn: https://www.linkedin.com/in/anbamo/
• GitHub: @Kamaranis