Table 2 Optimization hyper-parameters of different model for tuning the developed CP-EGBM.
From: An efficient churn prediction model using gradient boosting machine and metaheuristic optimization
Model | Function | Default value | Search space |
|---|---|---|---|
SVMRBF | \(\mathrm{C}\) | 1 | LB: 1E−1, UB: 1E |
Mapping of the feature space (\(\gamma\)) | 1/(#features) | LB: 1E−4, UB:1E4 | |
GBM | Number of estimators | 100 | LB: 100, UB:3000 |
Learning rate | 0.1 | LB: 1E−3, UB:1 | |
Maximum depth of DTs | 3 | LB: 1, UB: 10 | |
Minimum samples for split | 2 | LB: 2, UB: 10 | |
Maximum features | Sqrt(#features) | LB: 1, UB: #features | |
Sub-sample | 1 | LB: 0.5, UB:1 |
