Mohammad Abdollahi Data Scientist & Operational Researcher

Technical Lab · Synthetic demonstration

Technical Lab

Selected implementations, experiments and interactive examples in forecasting, optimisation and applied machine learning.

Synthetic demonstration · Independent educational implementation.

These pages show how a forecast is built, tested and turned into a decision. They are not the industry KTP system, the VN2 competition result, the doctoral simulations or the private ForecastOps project. Nothing here is a measured business impact.

The code was written for this site and lives in src/lib/lab/. It uses invented series, generic product IDs and small instances that can be solved in the browser. Course repositories on my GitHub account are not the source of these examples.

Forecasting

Horizons, intermittent demand and the evaluation design that keeps a forecast honest.

  • Multi-step forecasting

    Seasonal naive, recursive and direct forecasts for H1–H6, scored only on chronological origins.

    Synthetic demonstration

  • Intermittent demand

    A historical mean, classic Croston and TSB on synthetic demand with long runs of zeros.

    Synthetic demonstration

  • Chronological backtesting

    Move the forecast origin and see a random split place future weeks inside the training set.

    Synthetic demonstration

Inventory

The same synthetic decision as the interactive demo, read as a chain from forecast to order.

  • Forecast to inventory

    Expected demand, a coverage buffer, an order-up-to level and a pack-rounded order. The interactive demo uses this calculation.

    Synthetic demonstration

Optimisation

Small constrained problems, solved exactly and drawn so the binding constraint is visible.

  • Single-vehicle routing with time windows

    One depot, one vehicle and eight synthetic customers. Exhaustive search for this small instance, with capacity and delivery windows.

    Synthetic demonstration

  • Constrained procurement

    Pack size, a minimum order and a budget choose quantities for five generic SKUs.

    Synthetic demonstration

What is not here

There is no employer SQL, supplier rule, SKU list or private model. A larger dataset or a trained network would belong in object storage such as Cloudflare R2, not in this repository. Further examples can use the same page and module pattern.

Open the Forecast → Inventory demo if you want the full set of demand patterns and models. The lab page explains the decision chain without repeating that control panel.