🫀 ECG Arrhythmia Classification using Machine Learning

📌 Overview

This project implements and evaluates multiple supervised machine learning models to classify heartbeat signals from ECG (Electrocardiogram) data into arrhythmia types using the MIT-BIH Arrhythmia Dataset.

The project was completed as part of the assessment for the Principles of Data Mining and Machine Learning module (MOD 007892) at Anglia Ruskin University.

---

📄 Assessment Requirements & Implementation Summary

✅ Jupyter Notebook Tasks

Requirement	Implementation Summary
1. Download the Dataset	Downloaded from Kaggle MIT-BIH
2. EDA and Visualization	Data merged, null values checked, class distributions plotted, correlation matrix visualized
3. Pre-processing	Label encoding, feature-target split, normalization, class balancing suggestions noted
4. ML Algorithm Development	Implemented 3 classifiers: Random Forest, Support Vector Machine (SVM), Multi-Layer Perceptron (MLP)
5. Novel Contributions	Ensemble testing, performance metrics analysis, class imbalance analysis
6. Performance Comparison	Used Accuracy, Precision, Recall, F1 Score, Confusion Matrices

---

📘 Written Report Summary

Section	Summary
Introduction	Addresses the prevalence and fatal risk of arrhythmia and the need for accurate early classification using ML.
EDA & Data Visualization	Shows ECG signal graphs, heartbeat class distribution, and highlights dataset imbalance (majority class: normal)
Implementation	Describes full ML pipeline: data cleaning → correlation matrix → model building with math formulas
Results	MLP achieved highest accuracy (98%), followed by RF and SVM (97%). Performance metrics visualized and discussed per class
Conclusion	Suggested upsampling, label encoding for improved class balance and interpretability. Highlighted class imbalance as performance limiter.
References	Includes academic and industry sources for ML theory and ECG domain knowledge

---

📈 Key Metrics

Model	Accuracy	Precision	Recall	F1 Score
MLP (Neural Net)	98%	High	High	Highest
Random Forest	97%	High	High	High
SVM	97%	High	Moderate	Moderate

---

🧪 Machine Learning Models Used

• Random Forest Classifier (RF)
• Multi-Layer Perceptron (MLP)
• Support Vector Machine (SVM)

---

🧠 Dataset Classes

Class	Description	Label
N	Normal	0
S	Supraventricular	1
V	Ventricular	2
F	Fusion	3
Q	Unknown	4

---

📎 Links

• 🔗 Notebook: Arrhythmia Machine Learning.ipynb
• 🔗 Dataset: Kaggle – MIT-BIH Heartbeat Dataset
• 📄 Report PDF: Available on request or in coursework submission

---

🗃️ Submission Checklist (as per brief)

• ✅ Jupyter Notebook
• ✅ Written Report PDF
• ✅ Lab Logbook with GitHub link

---

📚 References

Key references include:

• MIT-BIH Dataset documentation
• Random Forests (Cutler)
• MLP (Prof. Alex Bronstein)
• SVMs (Tristan Fletcher)
• Performance metrics (Analytics Vidhya, Eugenio Zuccarelli)