Detection of antisocial personality disorder using optimal multi-head attention based auto encoder

Anti-social personality disorder (ASPD) is one of the mental disorders and the individual with ASPD generally violates the rules and becomes a criminal. Hence, it is essential to predict ASPD and provide proper treatment. This work presents a NLP (natural language programming) and optimal deep learning (DL) model that can automatically detect ASPD. This work undergoes major stages like pre-processing, optimal feature selection and DL based feature extraction and classification. In the pre-processing stage, the data imputation process is carried out. Then, the features essential for ASPD prediction are performed using the adaptive Marine Predators optimizer (AMPO). Finally, the selected features are predicted using the multi-head attention based autoencoder (MHA-AE). This classifier is used for predicting whether the individual is affected by the ASPD or not. The experimental analysis is carried out on the ASPD dataset and the proposed AMPO- MHA-AE model outperformed the conventional DL models. Accuracy and recall values achieved by the proposed ASPD model are 96.5% and 96.4% respectively. The results obtained are inspiring and could serve as a stimulus for researchers to enhance them, considering the significance and interest associated with this research direction.