Fig. 1

Simplified diagram of the hybrid model. The raw features provided with the datasets are input to the DNN. Real-valued features are input directly, and categorical features are input using a one-hot encoding. Each CNN block consists of two CNN layers followed by a max pooling layer. Features are extracted from a CNN layer. The CNN blocks are followed by a flattening layer and two dense layers leading to the classification. The labeled examples are used for training the network, but the network is not used for the final label prediction, as indicated by the dashed box around the DL classifier layers. Features extracted from the CNN layer are input to the XGBoost ML classifier for training and final predictions