Table 2 Hyperparameter search space for each algorithm.
Algorithm | Hyperparameters | Search space |
|---|---|---|
RF | n_estimators max_depth min_samples_split min_samples_leaf max_features | [100, 200, 500] [10, 20, 30] [2, 5, 7] [1,2,4] [‘auto’, ‘log2’] |
GB | learning_rate min_samples_split min_samples_leaf n_estimators | [0.1, 0.01, 0.001] [10, 20, 30] [2, 5, 7] [50, 100] |
MPR | Degree | 2 |
SVR | Kernel C | [‘rbf’] [5,7,10] |
MLR | n_jobs | 5 |
CatBoost | learning_rate n_estimators depth l2_leaf_reg | [0.01, 0.1, 0.2] [100, 200, 300] [4, 6, 8] [1, 3, 5] |
LightGBM | boosting_type num_leaves max_depth learning_rate n_estimators subsample_for_bin objective min_child_weight min_child_samples subsample | [gbdt] [31,63] [6,8,10] [0.01, 0.05, 0.001] [50, 100] [1000, 1500, 2000] [regression] [1,5,10] [20] |
ANN | activation optimizer loss epochs batch_size | [relu] [adam] [mean_squared_error] [50,100,150,200] [16,32,64] |
