How arpit9516/data-visualization-project works
Project Type and Purpose
• This is a full‑stack, no‑code AutoML platform built on Streamlit. It guides users through the end‑to‑end ML workflow—from data ingestion through model deployment as a REST API (via FastAPI). The UI is in Streamlit (app.py + pages/), the core pipeline logic lives in utils/, and models get stored/deployed.
High‑Level Architecture Components
a. User Interface (Frontend)
• Streamlit Web App (app.py + pages/01_…/09)
• Renders forms, charts, tables; captures user inputs and triggers pipeline steps.
b. Core Pipeline (Backend)
• utils/ modules implementing each step: data_loader, data_cleaner, data_profiler, data_visualizer, feature_engineering, model_selection, model_training, model_evaluation, model_explanation, model_deployment.
• Orchestrated by Streamlit callbacks in each page.
c. Data Storage
• Uploaded datasets (CSV/Excel) stored in ephemeral memory or temporary folder.
• Persisted models stored under a models/ or deployed_models/ directory (according to README).
d. Model Serving
• FastAPI service spun up by utils/model_deployment.py that exposes the trained model as REST endpoints.
• Automatic OpenAPI docs for the deployed model.
e. Third‑Party Libraries/Services
• SHAP or LIME for explainability (model_explanation).
• Plotting libraries (e.g., Plotly, Matplotlib) for visualization.
• Hyperparameter search frameworks (e.g., scikit‑optimize, Optuna) under model_selection.
Relationships and Data Flows
• User Browser → Streamlit App: requests pages, submits data
• Streamlit App → utils/data_loader: uploads & reads data
• Data flows sequentially through cleaner → profiler → visualizer → feature_engineering → model_selection → model_training → evaluation → explanation.
• Metadata/results (charts, tables, metrics) flow back to UI for display.
• Final trained model → utils/model_deployment → FastAPI server → external clients.
Key Architectural Patterns
• Layered MVC‑like structure:
• View: Streamlit pages
• Controller/Orchestration: app.py + page scripts
• Model/Business Logic: utils/ modules
• Pipeline pattern: each pages/0X_… step triggers a corresponding utils/ function.
• Microservice for model serving (FastAPI).