Glass Classification

Overview

A machine-learning classification project that predicts glass type categories from dataset features such as refractive index and elemental composition, comparing model accuracy and reviewing how well each model generalizes.

The problem

Manually sorting glass samples by type using raw chemical measurements is slow and inconsistent between observers. The goal was to see how reliably a trained classifier could replace that manual judgment using standard tabular features.

Key decision

Rather than reaching straight for the most complex model available, several classifiers were trained and compared side by side so that accuracy gains could be weighed against overfitting risk on a relatively small dataset — prioritizing a model that generalizes well over one that simply scores highest on training data.

Main features

• Dataset loading
• Data preprocessing
• Feature selection
• Model training
• Model evaluation
• Accuracy comparison
• Generalization analysis

Tech stack

Demonstrates: applied ML workflow, model evaluation, analytical reasoning.