What Muhammed demonstrated

Built a turbine energy-yield predictor end to end using leakage-free sensor inputs, a time-based held-out split, and both LinearRegression and RandomForest baselines. The RandomForest improved held-out MAE from 2.90 MWh to 1.53 MWh, with RMSE and R² reported on unseen later-period data. The model card clearly explains intended use, excluded downstream emissions, limitations, and a monitoring plan for drift.

• Leakage-Free Feature Selection
• Honest Held-Out Evaluation
• Residual-Based Error Reading
• Practical Model Monitoring