Self-Attention Mechanism and Balanced Loss Function Based Gated Recurrent Unit for Mental Disorder Detection System

The depression is a common mental health disorder that affected all around the world. The symptoms of depression such as sadness, angry, social behavior, sleep disorder and attempting suicide which are recorded from the World Health Organization (WHO). To detect the depression, sometimes classification such as depressed and not depressed are misclassified is challenged one. To overcome the problem, the research introduces the improved Self-Attention Mechanism and Balanced Loss Function based Gated Recurrent Unit (SA-BL-GRU) for mental health disorder detection system. The text and DAI- WOZ dataset contains a collection of information that provide as input to the proposed system. The collected input provided to pre-processing techniques such as data cleaning, tokenization, lemmatization and stemming to pre-process the data. The Bag of Words (BoW) is used for feature extraction phase to extract the significant features from the pre-processed data. The SA-BL-GRU is used for classification process that classify the accurate depressed and not depressed patients. The evaluation metrics such as accuracy, precision, recall and F1-score are used for estimate the performance of the evaluation metrics of 98.5%,98.3%,97.6% and 97.9% respectively.