Psychological Indicators Recognition Based On Behavioral Time Series Data

With the prominence of mental health problems,it is of great significance to identify individuals’ psychological indicators timely and effectively.However,the traditional methods of psychological measurement are intrusive and inefficient,which can not meet the needs of the current social mental health service.In recent years,with the development of the Internet and intelligent sensor technology,all walks of life have produced a huge amount of human behavior data.These data can capture fine-grained behavioral characteristics of individuals and provide a new perspective for research in various fields.Based on this,this study attempts to explore the relationship between behavioral time series data and psychological indicators,and further realize the automatic identification of individual psychological characteristics.This study first collects behavioral time series data from four scenarios:individual actions,(online/offline)interactive behaviors,and self-expression content.Specifically,in this study,Microsoft Kinect was used to collect individual gait data,MobileSens was used to collect individual smartphone behavior data,controller area network bus data logger was used to collect driving behavior data,and distributed crawler system was used to collect Weibo data.Then,using the time-frequency analysis method,this study extracts features that can represent the individual behavior pattern from these data respectively,and further realizes the analysis and recognition of the individual psychological indicators.The results showed that depression could be recognized based on gait behavior data(sensitivity=0.94,specificity=0.91,AUC=0.93).Based on the driving behavior data,the Big Five personality can be effectively recognized.The root mean square error of the regression model established on the five dimensions of the Big Five per-sonality is between 2.47 and 4.12,and the correlation coefficient can reach the level of strong correlation(0.56-0.88).Through the analysis of the mobile social behavior in smartphone,we found that depressed users received less calls from contacts and used social apps more frequently than nondepressed users.At the same time,based on the social behavior of the smartphone data model of machine learning can be implemented effectively for the depression of individual identification(sensitivity=0.76,specificity=0.25,AUC=0.71).In the self-expression scenario,through analyzing the Weibo texts of empty-nest youth and non-empty-nest youth,the results show that there are signif-icant differences in the psychological semantics and psychological indicators(positive emotions,negative emotions,etc.)of the two groups of Weibo users.The results show that behavioral time series data contain behavioral cues that can represent individual psychological indicators.Meanwhile,statistical methods and machine learning techniques can be used to effectively identify individual psychological indicators.Compared with the traditional psychological measurement methods,the psychological feature recognition method based on behavioral time series data has the characteristics of non-intrusive,traceability and automation.Therefore,the combination of this method and the traditional measurement methods can effectively improve the application range and measurement efficiency of psychological measurement.