model_v3/experiments.md

🏆 BEST PERFORMANCE MODEL (Lowest MAE)

python train_lstm.py --csv test1.csv --epochs 20 --batch_size 128 --hidden_size 256 --bidirectional --loss_type weighted_huber --model_out lstm_model_v3.pt

EVALUATION METRICS

MAE (Mean Absolute Error): 4.1815 mph RMSE (Root Mean Square Error): 7.6657 mph MAPE (Mean Absolute Percentage Error): 8.20% R² (Coefficient of Determination): 0.4294

Speed Range Performance: Low ≤30: 11.5730 mph (n=1889) Medium 30-60: 7.1722 mph (n=15717) High ≥60: 3.2165 mph (n=63171)

Saving predictions to test_predictions.csv Predictions saved with 80539 rows