Table 2 Key parameters used in model training and their set values.
From: The impact of CNN MHAM-enhanced WRF and BPNN models for user behavior prediction
Parameter name | Parameter description | Set value |
|---|---|---|
n_estimators | The number of decision trees in WRF | 100 |
max_depth | Maximum depth of decision tree | 10 |
min_samples_split | Minimum number of samples required to segment internal nodes | 2 |
min_samples_leaf | Minimum number of samples required for leaf nodes | 1 |
max_features | The maximum number of features to consider when finding the best split | “auto” |
criterion | The standard used when building the tree, “gini” or “entropy” | “gini” |
learning_rate | Learning rate in the training process of BPNN | 0.001 |
hidden_layers | The number of hidden layers in BPNN, in the format of “number of layers _ number of nodes” | “2_50” |
epochs | The number of iterations in the training process | 200 |
batch_size | Number of samples used in each iteration | 32 |
activation_function | Activation function for hidden layer and output layer | “relu” |
optimizer | An optimizer for updating the model weights | “adam” |
dropout_rate | Dropout ratio for regularization to prevent over-fitting | 0.2 |
weight_decay | L2 regularization term, which is used to control the model complexity | 0.001 |
