Lidar Regression with MLflow

Problem & context

Problem

Compare regression approaches for lidar observations and package the selected model for reproducible prediction.

Intended users

Technical reviewers evaluating modelling decisions and prediction-service behaviour.

Role and contributions

Regression modelling, experiment tracking and API implementation

Dataset / inputs

Lidar observations documented in the repository, with 45 observations reserved for final testing.

Methods & architecture

Regression models were compared with tracked experiments. The selected spline model uses 10 knots and regularisation alpha 0.01.

Implementation

1. Prepare the modelling data
2. Compare candidate regression models
3. Select the penalised spline configuration by cross-validation
4. Track runs with MLflow
5. Save test predictions
6. Verify the local prediction API

Verified results

Metrics below are reported under the project's documented evaluation conditions.

Limitations & next steps

A future deployment can be evaluated once hosting, monitoring and service-level requirements are defined.

Attribution & licence

See the repository for current attribution and licence information.