Results
Current research & experimental status
Proposed, implemented, verified and experimentally observed are kept distinct. Nothing is presented as a finished finding unless it is one.
The research design, questions, hypotheses, datasets and methodology defined in Research_Proposal.md v2.0.
The Python pipeline is scaffolded and running: forecasting model hierarchy, evaluation metrics, inventory simulator (lost sales) and a Streamlit prototype, with an automated test suite.
Software behavior is verified: the test suite passes and the pipeline is deterministic. This verifies the engine, not the research findings.
No final experimental results exist yet. Findings will appear here only after the M5 and Favorita experiments complete.
Research status
Where the project actually stands.
What is proposed, what is implemented and verified, and what experiments are still to be run — from the real repository state.
Proposal
- Research question defined
- Research objectives defined
- Dataset strategy defined
- Methodology defined (v2.0)
Implementation
- Forecasting pipeline scaffolded
- Inventory simulator implemented (lost sales)
- Evaluation metrics implemented
- Streamlit prototype implemented
- Automated tests implemented
Experiment
- M5 full experiment
- Favorita replication
- Cost-sensitivity sweep
- LLM evaluation
- Final statistical testing
The Experiment items are not yet complete. The website reports them as planned — nothing is presented as finished research.
Verified synthetic sanity check
A first, honest illustration — not final evidence.
Label: Verified synthetic-data sanity check — not final research evidence.
On the synthetic run, SARIMA had the lowest MAE, but XGBoost produced the lowest total inventory cost — a first illustration that the most accurate forecast is not automatically the most operationally valuable one.
Source: results/experiment_results.csv · 4 synthetic demand series × 7 models, 7-day horizon
SARIMA — lowest MAE
XGBoost — lowest inventory cost
Results will appear here after the final M5 and Favorita experiments. The values above come from a synthetic-data run used only to sanity-check the pipeline — they illustrate the research thesis but are not evidence.