MD Mizanur Rahman

Work · Forecasting

Mobility demand forecasting

  • Demo ready
  • Synthetic demonstration
  • Companion to coauthored research

Hourly zone demand from public NYC taxi records, a seasonal-naive baseline against gradient boosting, rolling-origin evaluation and labelled hypothetical scenarios.

Repository on GitHub

Problem

Random train/test splits make a forecaster look far better than it is. Every evaluation here is forward in time.

Contribution

Relationship to research: Companion to coauthored research. Independent follow-on to the coauthored Uber/Lyft study using public TLC data; not a replication of that article's observations.

Account owner’s role: Commissioned, reviewed and published the implementation. The code was produced with AI-assisted generation on commission and reviewed before release; it is not the original code of any cited publication.

What was built

  • Parquet adapter with schema validation
  • Lag features computed strictly from the past
  • Rolling-origin evaluator with errors by time block and zone
  • Scenario controls labelled hypothetical

Method

Seasonal naive uses the same hour one week earlier; gradient boosting uses 1h, 24h and 168h lags plus calendar features.

Data and access

NYC TLC yellow taxi monthly parquet files obtained by the user; synthetic four-zone fixture for CI.

Evaluation protocol

MAE and RMSE by time block and zone with sample counts.

Results

No measured result is published for this project. Demonstration outputs from the repository’s own synthetic fixtures are labelled as such in the repository and are not presented here as results.

Limitations

  • No Uber or Lyft proprietary data; no causal or revenue claim from scenarios.

Reproducibility

Offline synthetic smoke; the splitter is unit-tested for lookahead.

Repository metadata: Python, MIT, last push 2026-09-13, 0 stars.

Source publication and links