Supervised Learning on the Trojan Horse Dataset

This project implements a full supervised machine learning pipeline to classify instances in the Trojan Horse dataset.
The notebook walks through data preparation, model training, evaluation, and visualization using multiple ML algorithms.
The goal is to compare classifier performance and understand which input features drive accurate predictions.

---

Project Overview

This repository demonstrates a complete end-to-end ML workflow that includes:

• Data cleaning and preprocessing
• Handling categorical & numerical features
• Training and evaluating several classification algorithms
• Generating confusion matrices and performance comparisons
• Visualizing model behavior
• Interpreting feature importance (tree-based models)

This project is designed to show practical ML engineering skills using real datasets and reproducible workflows.

---

Pipeline Overview

1. Load & Clean Data

• Import Trojan Horse dataset
• Handle missing values
• Encode categorical features (One-Hot / Label Encoding)
• Normalize numerical input features (StandardScaler / MinMaxScaler)

2. Train/Test Split

• Hold-out split for unbiased model evaluation
• Ensures reproducibility with random states

3. Model Training (Classifiers Used)

The following models are implemented and benchmarked:

• Logistic Regression
• Decision Tree
• Random Forest
• k-Nearest Neighbors (k-NN)
• Support Vector Machine (SVM)
• Naïve Bayes
• Gradient Boosting / XGBoost
• Simple Ensemble / Voting Classifier

This allows comparison between linear, tree-based, distance-based, probabilistic, and boosting-based methods.

4. Evaluation

• Accuracy, Precision, Recall, F1-Score
• Confusion matrices for every classifier
• ROC curves (optional)
• Feature importance for tree-based models
• Misclassification analysis

5. Visualization

• Model performance bar charts
• Confusion matrix heatmaps
• Decision boundary visualizations (for 2-feature slices)
• Feature importance plots

---

Example Results Summary

Model	Accuracy	Notes
Logistic Regression	~53.87%	Baseline
Random Forest	~76.69%	Best overall, stable performance
Decision Tree	~64.90%	Good performance
k-NN	~67.17%	Sensitive to scaling
Naïve Bayes	~51.24%	Fast but innacurate due to correlated features
Ensemble	~68.26%	Competitive and stable

(Replace X% with your actual results if desired.)

---

Tech Stack

• Python
• scikit-learn
• Pandas
• NumPy
• Matplotlib / Seaborn
• Jupyter Notebook

---

Skills Demonstrated

• End-to-end supervised ML workflow
• Data preprocessing and feature engineering
• Training & comparing multiple classifiers
• Model evaluation and interpretation
• Visualization of ML outputs
• Reproducible Jupyter Notebook analysis

These skills directly map to real-world ML engineering and data science roles.

---