Research on Mental Disease Diagnosis Method Based on PMFS

Currently, the diagnosis of mental illnesses such as Alzheimer's disease, Schizophrenia, and Depression heavily relies on questionnaire surveys and the experience of doctors, resulting in a high rate of missed and misdiagnosed cases. In order to address this issue, a psychiatric disease diagnosis model based on parallel multi-layer feature selection (PMFS) is proposed. Firstly, based on the feature's inherent information, redundancy between features, and correlation between features and the learner, three-layer feature selection is performed in parallel using the Variance Thresholding, Max-Relevance and Min-Redundancy(mRMR), and Support Vector Machine-Recursive Feature Elimination(SVM-RFE), respectively, resulting in three feature subsets. Then, a voting score is assigned to these three subsets, and features with two or more votes are retained as the final feature set. Finally, the features in the set are taken as the input of the classifier, which is trained and tested to achieve the purpose of early diagnosis of mental illnesses. Experimental results demonstrate that compared with the SVM and RF classification models without using the PMFS algorithm, the PMFS-SVM and PMFS-RF models show better performance in terms of accuracy, sensitivity, recall, F1, and AUC.