Freelance data scientist · All case studies

A 100-metre water table, predicted continuously.

A GradientBoostingRegressor with nested spatial cross-validation mapped static water level across Sidama. Climate and environmental grids, hydrogeology codes, and binary thickness flags became a continuous 100 m raster the field team could actually use.

Client: Sidama groundwater programme. Built by Dilshad Raza.

Static water level in Sidama was known where someone had measured a well. Between those points the map went blank. Climate, environmental grids, and hydrogeology polygons lived in different files — and a random train/test split would leak neighbours into the score.

I trained a GradientBoostingRegressor with nested spatial cross-validation — five outer folds, three inner — so hyperparameters could not cheat on nearby wells. Climate and environmental grids joined hydrogeology polygons (HG_CODE). Regolith and quaternary thickness were recoded as binary predictors. The output was a continuous 100 m predicted SWL raster plus the table behind it.

Hire Dilshad Raza for similar freelance data science, machine learning, and AI automation in the UK, United States, and Australia.