Mental Disorders Prognosis and Predictions Using Artificial Intelligence Techniques: a Comprehensive Study

Mental diseases are a serious worldwide problem that requires precise and effective diagnosis techniques. This paper provides an in-depth overview and analysis of the latest developments in the application of deep learning techniques to identify mental illnesses. The key objective of this investigation is to evaluate the potential of DL algorithms in improving the precision and promptness of mental health diagnoses, which will lead to more focused and specific treatment strategies. The study discusses the present situation of mental health categorization and the limitations associated with traditional diagnostic methodologies. The dataset utilized in this study includes six types of mental health disorders: borderline personality disorder (BPD), depression, anxiety, bipolar, mental illness, and schizophrenia, and it is made up of 70,000 Reddit posts written by diverse people suffering from the mental disorder. The suggested study’s methodology makes use of text dataset standardization techniques to improve classifier performance. In the suggested investigation, we also discovered that min–max scaling enhances the performance of several classifiers by 26% to 40% compared to a conventional technique. We have utilized min–max scaling to transfer the features of the dataset to bring them to a similar scale. Later, we trained a total of eleven classifiers using the dataset and analyzed their performance using various performance measure parameters. The key findings are combined to demonstrate the ability of machine learning models to discriminate between various mental disorders, such as BPD, depression, anxiety, mental illness, schizophrenia, and bipolar disorder. In the presented study, we found that our proposed embedding GRU dense model achieves superior performance, demonstrating an accuracy of 92.59% and a training loss of 0.20. This performance surpasses that of conventional deep learning models, including LSTM, GRU, RNN, and other baseline models. Our experiments showed that the GRU dense model effectively captures complex patterns, leading to significant improvements in accuracy and loss metrics compared to widely used architectures. Based on the findings, the proposed classifiers can help medical professionals make more precise identifications of mental illnesses.