Skip to content
OrionStats
Log inStart learning
← Project Lab
PROJECT 25Advanced

Stock (Demand) Prediction

Business scenario

Buyers want a monthly unit forecast per product for the next quarter to plan purchase orders.

Problem statement

Forecast monthly units per product for Jan–Mar 2026 and evaluate accuracy with backtesting against simple baselines.

Learning objectives

  • Prepare a product-month panel
  • Compare naive, seasonal naive and regression/boosting forecasts
  • Evaluate with rolling-origin backtests

Instructions

Use inventory_monthly.units_sold or order lines. Backtest origins: 2025-06 … 2025-09 with 3-month horizon. Report WAPE overall and by ABC class.

Tasks

  1. Panel with lags and calendar features.
  2. Baselines.
  3. One ML model.
  4. Backtest and accuracy table.
  5. Final forecast and order-quantity suggestion.

Deliverables

  • Notebook
  • Forecast file
  • Accuracy report

Evaluation rubric

CriterionWeightExcellentProficientDevelopingInsufficient
Data & features20%Panel correct, leakage-free features.Correct and complete with minor gaps in rigour or presentation.Partly correct; important steps missing, unvalidated or unclear.Missing, incorrect, or not supported by the data.
Backtesting30%Correct rolling-origin evaluation vs baselines.Correct and complete with minor gaps in rigour or presentation.Partly correct; important steps missing, unvalidated or unclear.Missing, incorrect, or not supported by the data.
Model quality20%Sensible model, honest results.Correct and complete with minor gaps in rigour or presentation.Partly correct; important steps missing, unvalidated or unclear.Missing, incorrect, or not supported by the data.
Interpretation & recommendations15%Usable forecast and guidance.Correct and complete with minor gaps in rigour or presentation.Partly correct; important steps missing, unvalidated or unclear.Missing, incorrect, or not supported by the data.
Presentation15%Concise, well-labelled outputs a manager can act on without explanation.Correct and complete with minor gaps in rigour or presentation.Partly correct; important steps missing, unvalidated or unclear.Missing, incorrect, or not supported by the data.