Coffee Market Analytics
Reproducible time-series analytics pipeline for commodity price data, from ingestion and feature engineering to modeling and evaluation.
Repository · Details · PDF
Problem
Demonstrate an end-to-end, reproducible workflow for market data analysis that cleanly separates modeling from interpretation.
Approach
This project implements a reproducible time-series analytics pipeline for commodity price data. The workflow ingests public market data, stores it in a local SQLite database, engineers time-series features, trains a simple baseline model, and evaluates performance on a held-out test set.
A local LLM (via Ollama) is used strictly for interpretation and reporting, generating an analyst-style note based on model outputs and metrics — not for prediction.
Impact
Shows practical data science skills: structured pipelines, baseline modeling, transparent evaluation, and analyst-style reporting with LLM assistance.
Tools and Methods
- Python
- Time-series analysis
- SQLite
- Snakemake
- LLMs (Ollama)