Studies on Mental Health Recognition Based on Data Mining and Twin Learning Network

Abstract

This study was envisaged to develop a recognition method based on the data mining and twin network deep learning, in view of the recognition problems of the mental health data. Initially, the survey dataset was preprocessed using K-Means clustering and improved Apriori data mining methods. The Apriori data mining method was improvised, which significantly improved the pruning efficiency of the Apriori algorithm by introducing cumulative counting and address mapping tables. Subsequently, under the deep learning framework of the twin network, the reference dataset was included in the upper branch network and the survey dataset after clustering analysis, and data mining was included in the lower branch network. The upper branch network further integrated the channel self-attention mechanism, while the lower branch network further integrated the spatial self-attention mechanism. Based on various types of mental health data and reference datasets, identification experiments were conducted. The experimental results showed that the proposed method outperformed the Decision Tree (DT), Support Vector Machine (SVM), Convolutional Neural Network (CNN), and RNN methods using the four evaluation indicators of Precision, Recall, F1, and AUC. Furthermore, the developed method has higher pruning efficiency in data mining and higher accuracy in discriminating mental health.