How shafimohammad09/predicting-stroke-risk-using-machine-learning works
Project type and purpose
• This is a machine-learning pipeline with a lightweight Python web front end for stroke-risk prediction. The README and directory names show that the core is data ingestion, preprocessing, model training/evaluation, visualization, and finally a “front” folder serving predictions via a web API.
File structure
• Top-level “dataset” folder holds raw CSV.
• “main” folder contains orchestration scripts (main.py), custom helper functions, and visual reporting code.
• “milestoneX” folders break the pipeline into phases:
• milestone1.py – initial data exploration
• milestone2 – encoding & extra visualizations
• milestone3 – comparison of multiple models
• milestone4 – final pipeline adjustments
• “front” folder is a web application (front.py + webap.py) for serving the trained model.
• No external database; all data is file-based.
Key insights from README
• Preprocessing steps (handling missing BMI values, encoding categorical features)
• Modeling steps (Linear, Lasso, Ridge, Logistic regression)
• Visualization and evaluation metrics
• Future work (SMOTE, advanced algorithms, integration)
Main components
a. Data Layer
• dataset/data.csv (raw data source)
b. Preprocessing Layer
• Scripts in main/custom_function.py, milestone2/dataenco.py
• Tasks: missing-value handling, encoding
c. Feature Engineering & Pipeline Orchestration
• milestone scripts + main/main.py invoke each stage
d. Model Training & Evaluation
• milestone3/multimodelcomp.py + main/visual_main.py
• Implements regression and classification models, computes metrics
e. Visualization & Reporting
• main/visual_main.py, milestone2/visual_extra.py, presentation.pptx
• Generates distribution plots, precision-recall curves
f. Serving Layer (API & Front-end)
• front/webap.py – Flask (or FastAPI) service wrapping trained model
• front/front.py – front-end UI or CLI that calls web API
Relationships and interactions
• Front-end → API → Model Inference
Architectural patterns and design principles
• Linear pipeline (batch processing) for ML work
• Modular milestones for incremental development
• Separation of concerns: data, processing, modeling, visualization, serving
Technologies & libraries to call out
• Python (Pandas, NumPy, scikit-learn, Matplotlib/Seaborn)
• Flask or FastAPI for webap.py
• CSV file for data storage (no RDBMS)