Research On Psychological Testing Method Based On Negative Surveys

The continuous popularization of Internet technology has brought great changes in the development of traditional industries,it not only solves the space limitations of conventional information circulation and improves the work efficiency of ordinary users,also brings the risk of private data leakage and reduces the cost of criminals.In order to ensure the secure dissemination of sensitive data,a method of securely collecting private data is urgently needed.In traditional psychological testing,the investigators must completely collect the accurate results of each participant about the whole questionnaire and then summarize all the data for result analysis,and finally obtain the overall distribution information of participants,which may easily lead to data leakage from the collection and summary processes.Negative survey is a sensitive data collection method of a wide range of application scenarios,it can protect the private data of users while ensuring the overall distribution availability of survey data.But negative survey requires a reconstruction algorithm to estimate the distributional results of positive survey.In the psychological test questionnaire,the investigator assigns a score to each option,and the participants’ psychological state is assessed based on aggregate scores of the questionnaire.Although the existing negative survey algorithms can reconstruct the overall distribution of participants,it cannot reconstruct the total questionnaire score of each participant,which may lead to insufficient accuracy of the overall distribution of the reconstruction results or even errors in the actual survey results of the participants.In order to solve the problems mentioned above,this thesis designs a negative survey model based on the aggregate scores of the questionnaire and a corresponding reconstruction algorithm.Overall,the focus of this this thesis mainly involves the following three aspects:(1)A multi-question negative survey model based on aggregate scores was designed,which can be used to collect psychometric data.Compared with existing negative survey models,this model is mainly for negative survey model with aggregate scores,and it can be used for negative surveys that include multiple questions(conventional negative survey models are only suitable for negative surveys of a single question,or negative surveys that only consider small-scale problems).The main steps of the survey are as follows: first,the participants participate in the positive survey of each question,then aggregate scores of the questionnaire are calculated,model according to the survey of the participants in the selection results again randomly generated negative choice,and then the negative choice and aggregate scores information uploaded to the server in order to protect privacy,and finally,the server returns the questionnaire feedback of each participant according to aggregate scores information.(2)A negative survey reconstruction algorithm for reconstructing the overall distribution of psychological test data is proposed: NSto PS-S,and the algorithm is mainly based on generating functions,Bayesian formula,and grouping strategy.Since aggregate scores information of the questionnaire is required to be accurate,but NSto PS-II cannot accurately reconstruct the information,this chapter chooses to directly use the aggregate scores to optimize the reconstruction algorithm(named NSto PS-GF).In view of the fact that NSto PS-II cannot accurately reconstruct aggregate scores of the questionnaire,we use the generating function to combine aggregate scores of the questionnaire and negative selection to optimize the reconstruction algorithm(named NSto PS-GF).However,NSto PS-GF has poor reconstruction effect on negative data with medium scores.To solve this problem,this thesis divides the participant data into two sets according to the priority,and use NSto PS-II and NSto PS-GF for reconstruction respectively(named NSto PS-S).Experiments on the public DASS-42 scale dataset show that the average option error of NSto PS-S is about 0.8% lower than NSto PS-II.(3)Aiming at the prior knowledge in the psychological test questionnaire,a negative survey model with background knowledge and two reconstruction algorithms based on adjustment strategies are proposed.In conventional psychological test scenarios,investigators usually have some background knowledge in negative surveys,both NSto PS-BS and NSto PS-BS2 use the reconstruction result of NSto PS-S as the initial value,in which NSto PS-BS makes an overall evaluation on the data of each option in the initial value,if it does not meet the background interval,the current option will be adjusted,and the difference will be scaled to other undetermined options.NSto PS-BS2 determines the whole but only adjusts the reconstruction results of NSto PS-II in the grouping strategy.The experimental results on the DASS scale dataset show that when the number of participants is 200,the average option error of the reconstruction results of NSto PS-BS and NSto PS-BS2 is 3.83% and 3.43% lower than of NSto PS-S.However,as the number of participants increases,the reconstruction accuracy of NSto PS-BS is higher than that of NSto PS-BS.