| With the rapid change of modern society,mental health problems have become an increasingly prominent social problem,among which the incidence of depression disorder increases year by year,has become a prominent psychological problem in the whole society.As a special group,soldiers are more likely to suffer from mental health problems due to the characteristics of closed environment,heavy training intensity,strict discipline requirements and high-risk factor.They have long been under tremendous work,life and psychological pressure.Data show that in recent years,psychological disorders represented by depression disorders have become the new morbidity of our army,seriously endangers and weakens the combat effectiveness of our army.At present,a new round of military corrections is being carried out in depth.The scale and structure of the armed forces are being restructured and the force system is being reshaped.The standard requirements for war preparation are becoming higher and higher,and the scale and intensity of exercises,exercises and training continue to increase.At present,the psychological screening of conscription is mainly to screen and eliminate the young conscripts at high risk of schizophrenia,but the detection and identification of the high-risk population of depression disorder is still lack of effective means.Therefore,in the draft psychological test or military physical examination in detecting,increase the disorder of depression screening of highrisk population,depression people at high risk for block from the source,to reduce the incidence of troops depressive disorder,to promote the officers and men of mental health and to protect the camp safe and stable and improve the battle effectiveness is of great significance.The purpose of this study is that,in under the guidance of the concept of multiple mass fusion,the fusion of psychological test,analysis of natural language,eye patterns to explore a variety of technology and machine learning classification,depression high risk population screening scale for preparation,application of natural speech segmentation technology assessment unit scale of the structure of language cognition,answers depression risk groups in scale for in-depth analysis in the process of eye movement characteristics,and adopt the method of machine learning,respectively select the part of speech,reading sequence,scale response for the eye movement characteristics of the different cognitive process,this study provides experimental basis and data support for the exploration of psychological detection multi-quality fusion technology research and accurate and efficient screening of high-risk population with depression disorder.This research is mainly divided into three parts:In the first part,the development and reliability and validity test of the screening scale for high-risk population with depression disorder were conducted,including 3 experiments.The first experiment was to develop a screening scale for high-risk population of depression disorder.The second experiment was to test the reliability and validity of the screening scale for high-risk population with depression disorder.The third experiment used the method of natural language analysis to compare the screening scale for high-risk population of depression disorder with the traditional classical scale and analyze the language characteristics of the screening scale for high-risk population of depression disorder.In the second part,we studied the eye movement characteristics of high risk and normal people with depressive disorder scale.The fourth experiment was to study the differences of eye movement in the three interest areas of "question dry"(AOI1),"yes"(AOI2)and "no"(AOI3)between the two groups of people in the answering process of the scale,and extract more sensitive eye movement indicators for analysis,so as to explore the characteristics of eye movement of the two groups of people when answering the scale.Experiment 5 was a study on the relationship between subjective responses and objective eye movement indicators of the two groups of people.By analyzing the correlation between the scale scores of the two groups of people and their eye movement characteristics,the internal relationship was explored,providing a basis for machine learning classification of eye movement characteristics and auxiliary screening of high-risk groups of depression disorders.In the third part,the model of self-rating scale and eye movement feature fusion machine learning for screening high-risk population with depression disorder was explored.There are three experiments.In Experiment 6,the area of interest in answering questions was refined according to the part of speech,and the cognitive differences of the two groups of people were explored.The SVM-RFE classification algorithm was used to classify the groups,and the mode of using the difference of the frequency of the part of speech fixation to assist the identification of high-risk groups of depression disorder was discussed.In Experiment 7,eye-movement scanning paths of high-risk people with depression and normal people were compared in the process of answering the self-rating scale.Scan Match algorithm was used to classify the population,and the mode of assisted identification of high-risk people with depression based on scanning path features was explored.n Experiment 8,the eye movement difference between the positive and negative responses was explored question by question.The random forest algorithm was used to identify the positive and negative responses question by question,and the "question-by-question response eye movement feature Recognizer" was generated.SVM was used to construct the identification model of the high-risk population of depression disorder based on the "question-by-question scoring model".Results:Results 1.The Screening Scale for High-Risk People with Depression Disorder compiled in this study was composed of High Risk Depression-State(HRD-S)and High Risk Depression-Trait(HRD-T).The accessibility of items is 0.02-0.05;The total correlation coefficients of retained items were all greater than 0.40.The item discrimination degree was all greater than 3.00;Exploratory factor analysis showed that HRD-S included four dimensions,namely,emotional state,physiological state,cognitive state and sleep state.HRD-T contains six dimensions,namely pessimism and passivity,sensitivity and surrender,lack of interest,lack of confidence,impatience and anxiety,and vulnerability.Results 2.The reliability and validity of the Screening Scale for High-Risk People with Depression Disorders were tested.Cronbach’s alpha of HRD-S was 0.93,the split-half reliability was 0.85,and the retest reliability was 0.83.Cronbach’s alpha of HRD-T was 0.95,the split-half reliability was 0.88,and the retest reliability was 0.82.Confirmatory factor analysis on HRD-S table showed that CFI and TLI values were 0.98 and RMSEA value was0.03.Confirmatory factor analysis for HRD-T showed that CFI and TLI values were 0.98 and 0.99,respectively,and RMSEA values were 0.02.The follow-up study found that the prediction coincidence rate of the screening scale for high risk population of depression disorder was 88.69%.Results 3.The language structure characteristics of the Screening Scale for High-risk People with Depression Disorders were analyzed.HRD-S and HRD-T had the same characteristics as the classical scale(SDS,CES-D,BDI)in terms of the word frequency of high-frequency words(i.e.,the frequency of verbs,pronouns,nouns,adverbs and adjectives were more than those of other parts of speech)and the frequency distribution of each part of speech.The difference is that the word formation ratio of HRD-S is 60.50%,and the word formation ratio of HRD-T is 63.30%,which is slightly higher than the traditional classical scale.The average information ratio of HRD-S was 61.21%,and that of HRD-T was 61.73%.The difficulty of the questions was close to that of SDS and BDI.Results 4.On the AOI1 of the forward scoring questions,the average number of fixation points in the high-risk group was significantly higher than that in the normal group(p<0.01),and the average duration of fixation was significantly shorter than that in the normal group(p<0.01).On AOI2,the average number of fixation points in high-risk group was significantly higher than that in normal group(p<0.01),and the average fixation duration was significantly longer than that in normal group(p<0.01).On AOI3,the average number of fixation points and the average duration of fixation were significantly shorter in the high-risk group than in the normal group(p<0.01).As for the reverse scoring questions,the eye movement characteristics of the high-risk group with depression disorder were consistent with those of the forward scoring group in AOI1,while there was no significant difference in AOI2 and AOI3 from the normal group.Results 5.The scale score of high-risk population with depression disorder was negatively correlated with the number of fixation points on AOI3 and the mean fixation duration(r=-0.33,p<0.05;r= 0.33,p<0.05);The total score of normal population was significantly positively correlated with the number of fixation points on AOI2 and the average fixation duration(r=0.35,p<0.05;r=0.37,p<0.05).Results 6.The fixation frequency of verbs,pronouns,nouns,adjectives and numbers in the high-risk group and the normal group was significantly higher than that in the normal group(all p<0.05).The classification accuracy of the high-risk group and the normal group was 81.00% based on the difference of the fixation frequency of the part of speech.Results 7.When the scanning path was used to identify the two groups of people,the average classification accuracy of the single-track questions for the two groups of people was between 75.00-85.00%,and the population recognition rate of each dimension was between 70.00-80.00%.The recognition rates of pessimism and passivity,lack of selfconfidence,impatience and anxiety,self-blame and self-guilt,sensitivity and concession,and mental fragility were 78.18%,79.29%,72.00%,74.17%,77.35%,76.33%,respectively.The average population classification accuracy can reach 80.73%.Results 8.With the positive and negative answers as the classification labels,the eye movement characteristic patterns of different answers could be extracted,so as to further classify the choices of each question.The average recognition accuracy of 62 questions can reach 80.47% by using the eye-movement index of question by question response,and the crowd recognition rate constructed by "question by question scoring mode" can reach88.40%.Research Conclusions:The reliability and validity test of the first part shows that the scale has good reliability and validity.In addition,by using natural language word segmentation technology to the high-risk population screening scale for depression and traditional scale of the language structure comparison,this study compiled the scale and the traditional scale similar sentence structure,the degree of difficulty,but in the description on the level of emotion attitude and express feelings,language expression is more delicate,more suitable for combination of eye movement for depressive disorder high-risk groups on the syntactic structure of space and time characteristics.In the second part,from the temporal and spatial dimensions,it is proved that the high risk population and the normal population have their own eye movement characteristics when answering the scale.Similar to the patients with depression,the high-risk group of depression disorder showed the characteristics of reading the questions for several times in a short time and capturing less information in a single time,and showed a higher degree of negative cognitive involvement when answering the scale.At the same time,there was a correlation between the characteristics of eye movement and the scale scores,indicating that eye movement can be used as an indicator to distinguish different groups of people.The third part,combined with the different cognitive processes of the scale response process,respectively discussed three "scale + eye movement" screening high-risk population of depressive disorder.The identification rate of the eye movement features corresponding to the cognitive process in various screening modes can reach about 80.00%on the premise of small samples,and the identification rate of the crowd modeled by "pointby-question scoring mode" can reach 88.40%,reflecting the application value of eye movement features.The innovation and significance of this study are as follows:1.The paper discusses the theoretical structure of the status and characteristics of highrisk population with depression disorder,and uses natural language analysis and other techniques to compile the "screening scale for high-risk population with depression disorder",which provides a practical tool for psychological selection of recruits in the army;2.It was found that there was a high correlation between the eye movement characteristics and subjective response and objective indicators of the depression high-risk population scale test,which confirmed that the eye movement track characteristics were an effective indicator to distinguish the tendency of subjective response of psychological activities;3.To build "behavioral response-the part of speech" and "behavioral response-scanning path","whether response-eye movement features of recognition model",put forward " behavioral response,eye movement characteristics and sure-negative tendency" of the three-dimensional structure,improve the efficiency of the recognition of false negative and false positive,solved the problem of the self-report questionnaire response bias. |
PDF Full Text Request