Decision Tree Classifier from Scratch

This repository contains an implementation of a Decision Tree Classifier in Python from scratch, developed as part of the Advanced Programming course. The project includes both the algorithmic implementation and statistical analysis of the dataset.

Dataset

• Name: Airplane.csv
• Size: 103,904 samples × 24 features
• Target variable: satisfaction
• Source: Public dataset
• Preprocessing and exploratory data analysis (EDA) are included in the docs directory.

Features

• Custom Decision Tree implementation supporting:
  • Splitting criteria: Gini index & Information Gain
  • Hyperparameter tuning via coarse-to-fine search
  • Metrics: Accuracy and F1-Score (weighted 0.7 and 0.3 respectively)
• Post-pruning capability to reduce overfitting
• Visual tree export using Graphviz (saved as .svg)
• Statistical dataset interpretation and visualizations (via seaborn, matplotlib)

Requirements

• Python 3.x
• numpy, pandas, scipy, scikit-learn, matplotlib, seaborn, graphviz

Output

• Detailed performance metrics for train/validation sets
• Visual representation of the final decision tree
• Accuracy trends for different hyperparameter configurations

---

This project was created to deepen understanding of machine learning algorithms by building them entirely from the ground up, without relying on pre-built models.