Analysis of Single-Pilot Intention Modeling in Commercial Aviation

Table 1

Analysis of the network architecture of the single-pilot intention model.

S/N	Name	Type	Activations	Learnables	Description
1	sequenceInputLayer (sequence input with 92 dimensions)	Sequence input	92	—	Input sequence data into the network.
2	bilstmLayer (BiLSTM with 32 hidden units)	BiLSTM	64	InputWeights; RecurrentWeights; Bias	Learn bidirectional long-term dependencies between operation sequence time steps.
3	fullyConnectedLayer	Fully connected	11	Weights; Bias	Create a fully connected layer with 11 neurons for the 11 flight intent labels.
4	softmaxLayer	Softmax	11	—	Applies the softmax function to calculate the probability scores for each flight intent label.
5	classificationLayer	Classification output	11	—	Output the classification outcome according to the probability scores.