Table 3 Hyperparameter settings for the LSTM multi-attention model.
From: A hybrid super learner ensemble for phishing detection on mobile devices
Hyperparameter | Value |
|---|---|
Maximum sequence length | 75 |
Vocabulary size | \(|word\_index| + 1\) |
Embedding dimension | 100 |
Number of CNN filters | 64 |
CNN filter sizes | [3, 4, 5] |
LSTM units | 128 |
Number of attention heads | 4 |
Attention key dimension | 32 |
Dropout rate | 0.2 |
Loss function | Binary cross-entropy |
Optimizer | Adam |
Train-test split | 80–20% |
