How nclsprsnw/oasis works
Project type and purpose
• This is a full-stack machine-learning application for real-estate price prediction with a special focus on climatic events.
• It encompasses data ingestion and processing, model training and experiment tracking, an inference API, and a customer-facing web UI.
File structure
• data/
• raw/ – unprocessed historical real‐estate and climate data
• processed/ – cleaned and feature‐engineered datasets
• notebooks/ – Jupyter notebooks for EDA, preprocessing, feature engineering, model experimentation
• models/ – serialized model artifacts (pickles, ONNX, etc.)
• mlflow/ – configuration and deployment files for the MLflow tracking server
• api/ – Python code (e.g. FastAPI + Uvicorn) exposing a prediction endpoint
• web/ – frontend code (e.g. React/Vue/HTML+CSS+JS) that consumes the API
• requirements.txt – project dependencies
• docs/ – images and documentation
• CONTRIBUTING.md / README.md – guidelines and project overview
README insights
• MLflow is used to track experiments, log metrics, parameters, and register production models.
• The API is the inference gateway; the web app interacts with it to show predictions.
• Future plans include additional event types and expanded data pipelines.