Breast Cancer Decision Tree

Project Overview

An analytical and predictive application designed to help identify breast cancer markers using a custom-built machine learning decision tree algorithm. The program processes clinical dataset attributes, establishes logical classification splits, and predicts cancer classification with statistical accuracy reports.

Key Features

• Recursive entropy-based information gain computations for tree splits
• Data cleaning pipeline parsing CSV reports and resolving missing/null attributes
• Cross-validation testing evaluating prediction precision, recall, and F1-score
• Visual decision path generator showing logical branches created by the model

Technical Challenges & Solutions

Designing dynamic tree nodes without bloated memory allocations. Resolved by writing a strict pointer-based node manager in C++ that recursively prunes subtrees using custom threshold metrics, reducing memory footprint by 40% and preventing stack overflow errors on large training runs.