How ankuranks/digital_challenge works
Project type and purpose
• This is a data-science challenge repository, not a full-stack web app. It consists of three logical layers:
• Raw data storage (“Data” folder)
• SQL-based data exploration and transformation (“Part 1 SQL”)
• Machine-learning modeling and reporting (“Part 2 ML”)
• The goal is to take raw CSVs, run SQL queries to explore/transform, then build, evaluate, and explain ML models with Jupyter Notebooks, producing SHAP values and Sweetviz reports.
File structure
• Top-level:
• Data/ – raw inputs and a PPTX deliverable
• Part 1 SQL/ – three .sql scripts and documentation
• Part 2 ML/ – a Jupyter Notebook, outputs (SHAP_Values.csv, SWEETVIZ_REPORT.html, scores.csv)
• No web server, no microservices: it’s a linear data pipeline.
README for additional insights
• Confirms three deliverable sections (Data, SQL, ML)
• No external API integrations or deployment scripts
• Dependencies implicitly: an SQL engine (e.g., SQLite or Postgres), Python libs in the notebook (pandas, scikit-learn, shap, sweetviz)
Main components and interactions
a. Components
• Raw Data Store (CSV files in Data/)
• SQL Engine (runs Query1.sql–Query3.sql against raw data)
• Transformed Data (intermediate result sets or derived CSVs)
• ML Environment (Jupyter Notebook in Part 2 ML)
• Model artifacts (trained model, SHAP values)
• Reporting tools (SWEETVIZ_REPORT.html)
b. Data Flow
1. Raw CSVs → SQL Engine
2. SQL Engine → transformed CSVs or database views
3. Transformed data → Jupyter Notebook for training/evaluation
4. Notebook → output artifacts (scores.csv, SHAP_Values.csv, HTML report)
c. Architectural patterns/design principles
• Linear ETL → Train → Report pipeline
• Clear separation of concerns: storage, transformation, modeling, reporting
d. Technologies/frameworks
• SQL (engine unspecified, assume SQLite/Postgres)
• Python/Jupyter (pandas, scikit-learn, shap, sweetviz)
• CSV/HTML as interchange formats