Application Of Machine Learning In Social Anxiety Assessment

Face is the most important emotional clues in interpersonal communication and social interaction,a lot of research found the abnormal processing to the emotional face in social anxiety individuals,which using the eye movement.and the abnormal processing characteristics may be an important reason lead to difficulties in communication.However,a large number of studies have focused on processing mechanism in social anxiety,there is almost no objective biological markers can classify and identify effectively to social anxiety individuals.Support Vector Machine(SVM)is a classification algorithm in the field of machine learning,which is mainly used to solve the classification recognition pattern and regression analysis.Can we use SVM to predict social anxiety? In this study,we will collect the eye movement behavior data of all the subjects,and predict the classification accuracy rating of the using SVM classification model for individuals with high and low social anxiety.Social anxiety individuals show unusual eye movement processing characteristics when using different face stimulus,compared to the control group.this study verified and compared the processing model under two different faces stimulus structures for social anxiety individual,and using the SVM predict the classification accuracy rating.Experiment 1 examined the attention characteristics to a single face stimulus for the social anxiety individual.115 subjects(55 in the high social anxiety group and 60 in the low social anxiety group)were screened by questionnaire measurement.The subjects were given a single face stimulus and allowed to watch it freely.The stimulus presentation time was 10000 ms.The experiment l results show that high social anxiety individuals scanning path length increases,total fixation times and total fixation numbers is decreases to the eyes,compared to low social anxiety individuals,and the overscanning and avoidance to the eyes present to all emotion faces,include Anger,disgust,fear,happiness,neutral,which the result support the BFOE model.In Experiment 2,we examined the attention characteristics to a pair of faces for the social anxiety individual.A total of 115 subjects were screened by questionnaire measurement to view the stimulate freely using a stimulus structure similar to the point-detection paradigm(i.e.,an emotional face was paired with a neutral face).The stimulus presentation time was 10000 ms.The results showed that compared with neutral face stimuli,individuals high social anxiety individuals showed less total fixation times and total fixation numbers in threatening stimuli,which supported the avoidance hypothesis.Experiment 3 used SVM to establish the model of eye movement behavior and social anxiety,and used the model to predict social anxiety.The eye movement behavior indicators collected in Experiment 1 were used as the data set for classification prediction.The results showed that the accuracy was 80%,the sensitivity was 75%,and the specificity was 86%.The eye movement behavior indicators collected in Experiment2 were used as the data set for classification prediction.The results showed that the accuracy was 67%,the sensitivity was 62%,and the specificity was 71%.This study through the eye tracking technology combined with machine learning methods,not only verify the processing characteristics of social anxiety individuals to the emotional faces,but also realize the classification and recognition of social anxiety individuals by SVM,to provide reference for the evaluation and training of social anxiety disorder in the future.