How shanidevra/heart-disease-prediction works
Project type and purpose
• This is a data-science / machine-learning pipeline for heart disease prediction, implemented in a Jupyter notebook (heart disease predition (1).ipynb) and powered by a CSV dataset (heart.data.csv).
• The goal is to preprocess patient data, select relevant features, train three classification models (SVM, KNN, Neural Network), and compare their accuracy and stability.
File structure
• Top-level files only:
• README.md – project overview, dataset description, modeling steps, evaluation metrics
• heart.data.csv – raw input data
• Jupyter notebook – orchestration of preprocessing, feature selection, model training, evaluation, and visualization
• No dedicated src/, lib/, or services/ directories—this is a self-contained analytical workflow.
Key insights from the README
• Data Preprocessing stages: missing-value handling, scaling, encoding categorical features, train/test split.
• Three machine-learning techniques: SVM, KNN, Neural Networks.
• Evaluation metrics: accuracy, precision, recall, F1-score, ROC-AUC.
• Emphasis on comparative analysis.
Components
a. Data Layer
b. Preprocessing Module
• Data Cleaner (missing-value imputation)
• Feature Scaler / Encoder
• Train/Test Split
c. Feature Selection
• Correlation matrix analysis
• Selected feature subset
d. Modeling Suite
• SVM Trainer
• KNN Trainer
• Neural Network Trainer
e. Evaluation & Comparison
• Metrics Calculator (accuracy, precision, recall, F1, ROC-AUC)
• Model Comparison Dashboard
f. Visualization / Reporting
• Plots in Jupyter (confusion matrices, ROC curves)
• Summary report export
Relationships and data flow
• Preprocessing Module → Feature Selection → Modeling Suite
• Each model trainer → Evaluation & Comparison
• Evaluation → Visualization / Reporting
Architectural patterns and design principles
• Pipeline pattern: a linear workflow of discrete stages (ETL → modeling → evaluation).
• Modular decomposition: each stage (preprocessing, feature selection, each model) is its own component.
Technologies and frameworks to annotate
• Python (language)
• pandas, NumPy (data manipulation)
• scikit-learn (SVM, KNN, metrics)
• TensorFlow / Keras or PyTorch (neural network)
• Jupyter Notebook (orchestration and visualization)