🩺 ECG Signal Classification using Machine Learning

I did this project using machine learning to classify ECG (Electrocardiogram) signals as normal or abnormal using real heartbeat data. It was inspired by the idea of using AI in predictive healthcare — helping detect irregular heart activity automatically using signal data.

---

Dataset Used

The project uses ECG datasets from the PTB Diagnostic ECG Database. The files used are:

• ptbdb_normal.xlsx – contains normal ECG signals
• ptbdb_abnormal.xlsx – contains abnormal ECG signals

Each row in these files contains 187 numerical values representing a patient's heartbeat signal.

---

Project Steps

1. Data Loading

• Loaded the normal and abnormal ECG datasets using pandas
• Each sample was labeled:
  • 0 → normal
  • 1 → abnormal

2. Preprocessing

• Added a label column to both datasets
• Combined both into a single DataFrame
• Shuffled the rows randomly for unbiased learning

3. Splitting the Data

• Used train_test_split() to divide the data:
  • 80% for training
  • 20% for testing

4. Model Training

• Trained a RandomForestClassifier using the signal data
• The model learned to detect patterns in ECG signals

5. Model Evaluation

• The model achieved 100% accuracy on the test set
• Evaluated using:
  • Accuracy score
  • Confusion matrix
  • Precision, recall, and F1-score

---

Results

The model's performance was perfect:

• Accuracy: 100%
• Precision, Recall, F1-score: All 1.00

Accuracy: 1.0

Confusion Matrix: [[ 787 0] [ 0 2123]]

Classification Report: precision recall f1-score support

       0       1.00      1.00      1.00       787
       1       1.00      1.00      1.00      2123

accuracy                           1.00      2910
macro avg       1.00      1.00      1.00      2910
weighted avg       1.00      1.00      1.00      2910

Screenshots

Normal vs Abnormal ECG Samples

Shows first 5 rows of normal and abnormal ECG data used for training.

Merged Dataset Overview

Combined dataset with over 14,000 samples and 310 features.

Final Model Evaluation

Model performance showing 100% accuracy and perfect classification report.

---

Requirements

Install the dependencies using:

pip install -r requirements.txt