A MENTAL ILLNESS EARLY WARNING ANALYSIS MODEL USING AN IMPROVED SVM ALGORITHM AND EEG SIGNALS

Abstract

The fast-paced lifestyle, intense competition, and constantly changing environment of modern society are important factors contributing to the emergence of psychological disorders. Factors such as work pressure, family conflicts, and interpersonal relationship difficulties pose challenges to individual mental health. Prolonged use of electronic devices and the widespread use of social media further exacerbate mental health issues, as the constant influx of information makes it difficult for people to relax and focus. The harm of psychological disorders to individuals and society cannot be ignored. It not only diminishes the quality of life for patients but also affects work efficiency and social participation. In extreme cases, psychological disorders can lead to serious consequences like suicide, causing immense pain to families and society. Therefore, prioritizing mental health, raising awareness of mental well-being, and timely intervention and treatment of psychological disorders are crucial for individual happiness and social stability. Based on this, this study employs an improved Support Vector Machine (SVM) and Electroencephalogram (EEG) signals for the diagnosis of psychological disorders, aiming for more effective treatment of mental health conditions. Specifically, the EEG signals of the experimental subjects are first collected as training data for the model. The dataset undergoes preprocessing operations such as cleaning and normalization. Due to the limited availability of published information about individuals with psychological disorders, an ensemble SVM algorithm is used as a warning model to improve the accuracy of psychological disorder diagnosis. This algorithm utilizes a differential measurement method to adjust the selection of SVM parameters. The training sample size remains constant, with several iterations of training performed. By combining the differential measurement method, a specific parameter combination is obtained, maximizing the differentiation between base classifiers. This ultimately addresses issues such as small sample size and insufficient sample information in the dataset. Experimental results demonstrate that the model used in this study exhibits good performance in diagnosing psychological disorders.