Fig. 3
From: Daily insider threat detection with hybrid TCN transformer architecture
Model Architecture. The model takes window-based data as input, with w \(\times\) 104, where w represents the number of time steps, and 104 is the feature dimension. After passing through four layers of TCNBlock, the data is fed into two layers of Transformer. The final output shape is w \(\times\) 1, corresponding to predictions for each time step.
