Intelligent identification of risk personnel psychological profiling system based on FMT features

Abstract

The present invention discloses a psychological portrait system for intelligently identifying risky personnel based on FMT features, which includes the following steps: S1: feature screening, key feature collection and risk feature screening; S2: psychological portrait drawing and evaluation of high-risk personnel; S3: scale compilation, test administration and data collection; S4: data analysis, determining the portrait model through a Binarylogistic regression model; S5: obtaining the final risk level of the tester based on a fuzzy comprehensive judgment method. The psychological portrait system for intelligently identifying risky personnel based on FMT features constructs a psychological portrait of high-risk personnel by collecting individual psychological feature information of suicides, wounding persons and other individuals who have engaged in risky behaviors. According to the fitting of the psychological portraits of the tester and the high-risk personnel, a fuzzy comprehensive judgment method is used to distinguish high-risk personnel from non-risk personnel, so as to find the real high-risk population in a more comprehensive, specific and targeted manner.

Description

Technical Field

[0001] The present invention relates to the field of psychological assessment, and in particular to a psychological portrait system for intelligently identifying risky personnel based on FMT features. Background Art

[0002] Psychological testing is a tool that measures a certain psychology or behavior of an individual according to a standardized procedure. The measurement results can be used to determine the differences between people. Psychological measurement is a technology that standardizes people's psychology and behavior. Chinese psychologist Peng Kaiping believes that "psychological testing is a scientific method that observes a few representative behaviors of people, and infers and quantifies the psychological characteristics that run through people's behavioral activities based on certain principles."

[0003] In recent years, mental health has become more and more the focus of modern people. It is on par with physical health and has become one of the key indicators for measuring human health. Actively preventing and reducing mental illnesses, curbing the occurrence of psychological crisis events, improving the quality of work and life, and promoting the harmonious development of individuals and society have become issues of common concern and research.

[0004] At present, the contradictions related to mental health needs are mainly manifested in the following aspects: the mental health service resources of hospital outpatient clinics and mental health centers are very limited and cannot fully meet the assessment, diagnosis and treatment needs of all patients; the traditional assessment tools used by psychological counseling and EAP agencies in their services lack a complete and effective assessment system; sensitive and fast early warning methods for college students' psychological crises still need to be vigorously developed; the instability of the mental health level and stress level of corporate employees and managers has greatly affected the talent reserve and production efficiency of enterprises, and has continuously caused serious problems such as staff loss and low employee satisfaction. Among these problems, the lack of a good mental health assessment system is the basis and key to all problems.

[0005] At present, the mental health assessment tools used in China include tools developed abroad and revised domestically, and tools independently developed domestically. The former accounts for the vast majority of all mental health assessment tools, but will cause various inadaptability due to cultural differences and translation problems; the latter is mainly used for special research, which is relatively simple and not systematic in content and structure, and requires a lot of in-depth exploration and development. There are several types of mental health assessment tools: ① Comprehensive assessment tools, mainly including the Symptom Checklist-90 (SCL-90) and the Minnesota Multiphase Personality Inventory (MMPI; including MMPI-1, MMPI-2 and MMPI-Youth Edition); ② Assessment tools for specific pathological symptoms, mainly including the Hamilton Depression Rating Scale, the Depression Self-Rating Scale, the Beck Anxiety Questionnaire, the Marks Obsessive-Compulsive Scale, etc.; ③ Clinical diagnostic systems, mainly including the Diagnostic and Classification Manual (DSM), etc. Most of the mental health research in China directly transfers relevant Western theories and concepts, and directly applies them to scientific research, psychological counseling and consultation by revising the scale. However, the content of mental health defined in the cultural and social environment of the East and the West is not exactly the same. The cultural and psychological connotations contained in Western questionnaires are difficult to accurately and effectively apply to the Chinese population even after revision. Therefore, it is urgent and important to build a mental health assessment tool that conforms to the Chinese cultural background and can provide personalized psychological assessment services.

[0006] At present, in the field of mental health assessment, the existing R&D and technical problems mainly include:

[0007] The tool system is seriously aging and outdated. The commonly used assessment tools in China lack the absorption of the latest research results of contemporary mental health. Most assessment institutions directly use traditional mental health assessment tools, such as SCL-90, or develop new scales by splitting and recombining the dimensions in different traditional scales. They do not keep up with the times and select the latest research results. They do not use the latest research results to measure the mental health status of contemporary people. In addition, the language of the assessment items uses a lot of expressions from many years ago, without considering the changes in people's language habits caused by the changes of the times, which can easily lead to ambiguity and make the assessment wrong. At the same time, it is also easy to make the subjects feel uncomfortable and produce resistance. At present, most of the mental health assessment tools used by most assessment companies are assessment tools translated from foreign countries earlier. The language expression is stiff, which often causes dissatisfaction among the subjects and leads to inaccurate assessment results.

[0008] The current scales are mainly based on symptom measurement. Most of the tools developed and used in China are mainly based on symptom measurement. They can only judge whether an individual suffers from a mental illness from a negative perspective, and cannot conduct a comprehensive and holistic assessment of mental health from positive perspectives such as the scope of social activities, life satisfaction, self-esteem, social support, and subjective well-being. Therefore, it is difficult to fully reflect the true level of individual mental health. In addition, the evaluation results of symptom-based measurements often require people with a certain foundation of psychological knowledge to be better understood. For ordinary test subjects or corporate managers, it is difficult to understand, and often requires professionals to interpret the evaluation results before they can fully understand them.

[0009] Symptom screening and functional assessment are confused. Existing mental health assessment tools do not clearly distinguish functional assessment from symptom screening. They usually regard functional assessment as an aspect of symptom assessment and do not realize the special role of functional assessment. For the assessment of the general population, the assessment of social function is more holistic and comprehensive, which is more reasonable and applicable. Just like patients who go to the hospital for treatment do not need to have X-rays, most people can get feedback through medical guidance and clinical interviews. In other words, the current mental health assessment tools mostly distinguish between normal people and people with psychological problems, but there is no better screening for those who are in "psychological sub-health", that is, although some people do not currently have obvious psychological problems such as depression and anxiety, they may be on the verge of psychological problems due to problems in psychological functions such as personality, coping style, and psychological resilience.

[0010] The etiology analysis framework of mental illness is not complete. The existing assessment tools provide rich materials for verifying the clinical symptoms of mental illness, but lack contribution to the etiology analysis of mental illness. More and more studies have confirmed that mental illness is related to a variety of internal and external factors such as individual defense mechanisms, attachment patterns, social support, and growth environment. However, the results of these studies have not been absorbed by the recent development of mental health assessment tools, and cannot provide etiology analysis references for the clinical treatment of psychological counselors and psychotherapists.

[0011] The quality of localized scales is low, and the quality of norms is questionable. Direct translation and use of foreign assessment tools will also have problems with measurement norms and text quality. Foreign tools must have domestic norms before they can be used. In addition, translated tools often have language and text quality problems, and there will be deviations in wording, grammar, and expression habits. These factors will significantly interfere with the evaluation effect.

[0012] The evaluation results are mostly based on the individual situation of the testees. At present, the evaluation results given by most evaluation agencies are used to evaluate the mental health level of the testees, but the companies or universities where the testees work are often the ones who need the evaluation results. Companies or universities are more concerned about what risks and losses their employees or students' mental health problems will bring to the company.

[0013] The evaluation method generally adopts the score conversion method. There are three main score conversion methods for evaluation: the first is the development scale. Comparing an individual's performance with people of various development levels can make a development scale. Commonly used are age scales and grade scales. The second is the quotient. A scale made by comparing a person's mental development level with his actual age, the most famous of which is IQ. In educational tests, quotients are sometimes used to indicate the rate of educational development or achievement. Commonly seen are educational quotients and achievement quotients. The third is percentile rank. The percentile rank indicates the relative position of an individual in the norm group. The fourth is the standard score. It is a scale that expresses the distance between the original score and the mean in units of standard deviation. Because its unit is standard deviation, it is called a standard score. There are two ways to convert standard scores. One is linear conversion, that is, using the formula Z = (XX) / SD for conversion, where Z is the standard score, X is the test subject's score, X is the sample mean, and SD is the sample score standard deviation. The second is to convert the original score into a percentile rank, and then look up the corresponding standard score from the normal curve area table. This conversion has a premise that the scores of the measured traits should actually be normally distributed. These score conversion methods usually require clear numerical definitions, but in fact, in daily life and work, we often encounter many fuzzy things without obvious quantitative boundaries, such as "beautiful, hot, far", etc. These concepts cannot be simply expressed by "yes", "no" or "number". Usually, our evaluation of a thing is not just from one aspect, but involves many indicators, and this is even more true for the mental health of an individual.

[0014] The quality of mental health is mainly determined by comparing with the norm. The norm-referenced score compares the test subject's score with a group of people with certain characteristics, and reports his score based on his relative position in the group. Here, the reference group used for comparison is called the norm group, and the score distribution of the norm group is called the norm. Determination of the norm group: The norm group is a group of people with certain common characteristics. When determining the norm group, we should pay attention to the following: first, the composition of the group must be defined; second, when the group is too large, the norm group must be a representative sampling of the group to be tested; third, the sampling process must be described in detail; fourth, the sample size must be appropriate; fifth, we should pay attention to the timeliness of the norm; sixth, we should combine the general norm with the special norm.

[0015] Therefore, through the existing methods, we can only judge the relative position of an individual in his or her group. Then, through the norm reference method, we can only conclude that when an individual's mental health score is far from the norm score, his or her mental health is poor. Therefore, the individual's mental health will change with the change of the reference norm, which makes it impossible to give a more accurate mental health status. Summary of the invention

[0016] The purpose of the present invention is to provide a system for intelligently identifying the psychological profile of risky personnel based on FMT features, so as to solve the problems raised in the above-mentioned background technology.

[0017] To achieve the above object, the present invention provides the following technical solutions:

[0018] The system for intelligently identifying risk personnel psychological portraits based on FMT features includes the following steps:

[0019] S1: Feature screening, key feature collection and risk feature screening;

[0020] S2: Model outline and evaluation of psychological profiling model for high-risk individuals;

[0021] S3: develop the scale, administer the test, and collect data;

[0022] S4: Data analysis, determining the profiling model through Binarylogistic regression;

[0023] S5: Based on the fuzzy comprehensive judgment method, the final risk level of the tester is obtained.

[0024] As a further solution of the present invention: the key features in S1 are collected by using a semi-structured interview method to interview 10 senior managers of enterprises in different industries for 30-45 minutes to understand employee risk events that have occurred in the enterprise, such as extreme events such as extreme psychological behavior, hurting people, and deliberate destruction, as well as some risky behaviors that enterprise managers are worried about employees. 56 key features of risky employees are sorted out, and the risk features in S1 are screened by voting to evaluate the key features. 8 senior HRs are invited to vote for each key feature, and more than 60% of the items related to the key features of risky employees selected by senior HRs are retained. Thus, a total of 34 risk features are selected, and 6 types of psychological risk employee character portraits are initially formed, including extreme psychological tendency psychological portraits, hurting tendency psychological portraits, resignation revenge psychological portraits, high energy consumption tendency psychological portraits, negative energy transmission psychological portraits, and glass bottle image psychological portraits.

[0025] As a further solution of the present invention: the psychological portrait model of S2 medium and high-risk personnel is mainly drawn by literature analysis method, and relevant domestic and foreign literature is read and studied on different psychological risk portraits, and the risk models of different character portraits are learned and mastered. By sorting out and summarizing the existing literature and academic achievements, we comprehensively outline 6 types of psychological risk character portraits from the five dimensions of external stimulation received by individuals, individual emotional state, personality, cognition and psychological resilience. The psychological portrait model evaluation of S2 medium and high-risk personnel includes three levels: employee social function status screening, clinical symptom verification and psychological motivation analysis. Employee social function status screening covers interpersonal quality, behavior efficiency, and emotional Status, positive feelings and scope of activities. Interpersonal quality assessment includes whether interpersonal relationships are stable and harmonious; efficiency assessment includes the operating efficiency of functions such as memory, reaction, and learning; emotional status assessment includes whether the subject can maintain a stable and peaceful emotional state; positive feeling assessment includes whether the individual has a sense of happiness and accomplishment, etc.; activity scope investigation includes whether the individual has enough rich social interactions and entertainment and leisure activities. Clinical symptom verification mainly analyzes employees' depression, hostility-aggression, paranoia and other psychological problems, so as to clarify the psychological problems through the clinical symptom verification scale. Psychological motivation analysis mainly includes stressful events, defense mechanisms, social support, etc., so as to analyze the psychological motivations of employees.

[0026] As a further solution of the present invention: the scale is compiled in the S3, and the initial scale is compiled according to the dimensional composition of different types of psychological portraits. At the same time, the Self-Assessment Scale of Mental Health Symptoms (SCL-90), the Extreme Psychological Tendency Scale in the Interpersonal Psychological Model, the Barratt Impulsiveness Scale, the Stress Perception Scale (PSS) compiled by Dr. Cohen, etc. are referred to, and a total of 450 original scales are compiled. The test and data collection in the S3 are carried out in the form of fixed investment, random investment and large-scale investment. Among them, the fixed investment will accurately distribute 6 copies to typical portrait characters, such as the subscale for measuring extreme psychological tendencies will be distributed to testers who have had extreme psychological behaviors but attempted extreme psychological behaviors, and the subscale for measuring the tendency to hurt others will be distributed to testers who have had hurtful behaviors, etc. Among them, the random investment will be randomly distributed to the general normal population, and the large-scale investment is mainly placed on the system APP platform for free measurement by users. Through data cleaning and elimination of invalid questionnaires, a total of 16,805 valid questionnaires were collected.

[0027] As a further solution of the present invention: the data analysis in S4 includes reliability and validity analysis. Reliability is an indicator that reflects the stability and identity of the tester. The "Cronbach's α coefficient" indicator is used to test the stability and identity of the scale. Validity refers to the extent to which a test measures the characteristics it is intended to measure, and is an indicator that reflects the authenticity and accuracy of the measurement. SPSS20.0 is used to perform exploratory factor analysis on the collected data, and the compiled scale is tested for validity. Inappropriate questions are deleted through factor analysis to obtain good structural validity.

[0028] As a further solution of the present invention: the portrait model is determined in S4, and the data of the testers who take different risk type scales are divided into two categories, namely, extreme psychological and non-extreme psychological, hurtful and non-hurting, etc., and the binary logistic regression method is used to find out the independent variables that really cause individuals to take extreme psychological behaviors, hurtful behaviors, quitting jobs for revenge, etc., and the log-linear model is used for statistical analysis. In the log-linear model, when its binary variables are used as dependent variables, the log-linear regression model becomes a Binarylogistic regression model. The following introduces the derivation and solution of the BinaryLogistic model in detail, taking the extreme psychological tendency model as an example.

[0029] Assume that the nth individual, n∈N, takes extreme psychological behavior with y n Indicates that y n =1 represents that the individual has adopted extreme psychological behavior, otherwise y n = 0, assuming that in theory, there is a continuous response variable y * n Represents y n The probability of occurrence ranges from negative infinity to positive infinity. If the value of this variable exceeds a critical point m, such as m = 0, the traveler n will take extreme psychological behavior, so:

[0030] When * n >0 time y n =1

[0031] When * n ≦0 when y n =0

[0032] If we assume that the response variable y * n The factors affecting the selection of independent variables, x ni There is a linear relationship between (i=1,2…,m), namely

[0033] y * n =β0+β1xn1 +…+β m x nm +ε j 1.1

[0034] Where: j is the error term, which obeys the Logistic distribution; β0 is the intercept, which is usually called the constant term; β i (i=1,2…,m) is x ni The partial regression coefficient of (i=1,2…,m) is as follows. From formula 1.1, we can know that the conditional probability of the nth individual taking extreme psychological behavior is:

[0035] P(y n =1 / x ni )=P[(β0+β1x n1 +…+β m x nm +ε j )>0]1.2

[0036] By calculating and solving, we can get:

[0037]

[0038] Where P(y n =1 / x ni ) is a nonlinear function composed of explanatory variables (influencing factors), which can be transformed into a linear function.

[0039] First, define the probability of extreme psychological behavior being selected by individual n as:

[0040]

[0041] Then the probability ratio of extreme psychological behavior being selected and not being selected is:

[0042]

[0043] This ratio is called the occurrence ratio of the event, abbreviated as Odds. From formula 1.5, we can see that the larger the p, the larger the odds; the smaller the p, the smaller the odds. In order to measure the degree of influence of a certain independent variable on whether the dependent variable takes extreme psychological behavior, we define OddsRatio, abbreviated as OR, or odds ratio, as follows:

[0044]

[0045] This means that when the influencing factors of other independent variable selection behaviors remain unchanged, the independent variable x ni Change of one unit, the OR corresponding to the dependent variable changes exp(β i). We used OR to conduct sensitivity analysis on the factors that influence individuals' choice of extreme psychological behaviors.

[0046] Since Binarylogistic regression is a nonlinear model, the maximum likelihood estimation is used to solve the parameters. Assume that there is a population consisting of N individual choice behaviors, Y1, ...YN, from which n cases are randomly selected as samples, and the observed values ​​are labeled y1, ...yn. Let p i =p(y i =1|x i ) is a given influencing factor x i The result y is obtained under the condition i =1, then y i =0 is p(y i =0|x i )=1-p i , so the probability of getting an observation is:

[0047]

[0048] where y i =1 or 0, because the observations are independent of each other, their joint distribution can be expressed as the product of the marginal distributions, as shown below:

[0049]

[0050] Formula 1.8 is called the likelihood function of whether n individuals will have extreme psychology, where:

[0051] In order to find the parameter estimate that can maximize the value of this likelihood function, * Performing logarithmic transformation, we get the following function:

[0052]

[0053] In order to estimate ln(L * ) reaches its maximum value when the population parameter β0…β m , find their partial derivatives respectively, and then set them equal to 0, and get the following formula:

[0054]

[0055] …

[0056]

[0057] As can be seen from the above formula, there are m+1 independent variables in the system of equations, and there are m+1 simultaneous equations to estimate their values. Since all of the above are nonlinear functions, manual calculation and solution are very difficult. This paper uses SPSS20.0 to solve them.

[0058] Through binary logistic regression, we established 6 regression equations to determine whether the tester belongs to a certain type of risk profile.

[0059] Mainly based on the equation derived from the regression, a certain weight is assigned to each measurement dimension. After the tester completes the test, our system background will automatically add up the dimension values ​​related to a certain portrait, and bring the value of each dimension into the equation of different portraits, and calculate the risk value corresponding to different categories of portraits. The closer the risk value is to 100, the more likely it is that the behavior of the person represented by the portrait will be adopted. Taking extreme psychological tendencies as an example, if the tester's extreme psychological tendency risk score is closer to 100, it means that he matches the personality, emotions, behavior and other characteristics of extreme psychological people, and is more likely to adopt extreme psychological behaviors.

[0060] As a further solution of the present invention: the fuzzy comprehensive judgment method in S5, since human psychology is often unpredictable, the level of individual psychological risk often involves multiple factors or multiple indicators. At this time, it is required to make a comprehensive evaluation of things based on these multiple factors, rather than evaluating things based on the situation of only one factor. Therefore, the judgment of comprehensive psychological risk mainly adopts the fuzzy comprehensive judgment method. Here, judgment means to evaluate and distinguish the pros and cons, good and bad of things according to given conditions; comprehensive means that the judgment conditions include multiple factors or multiple indicators; fuzzy means using fuzzy mathematical calculation methods for processing. Fuzzy comprehensive judgment can also effectively integrate different types of data, including non-normally distributed data. It has stronger adaptability and ultimately obtains an evaluation of the subject.

[0061] The internal logic of fuzzy comprehensive evaluation:

[0062] Fuzzy comprehensive evaluation is a very effective multi-factor decision-making method that makes a comprehensive evaluation of things affected by multiple factors. Its characteristic is that the evaluation result is not absolutely affirmative or negative, but is expressed as a fuzzy set.

[0063] Membership degree or membership function is the basis of fuzzy mathematics and fuzzy systems. The concept of "membership degree" is introduced to describe the intermediate transition of differences. It is an approximation of accuracy to fuzziness. Its mathematical definition is:

[0064] For a fuzzy set A on the universe U, specify a mapping from U to [0,1]:

[0065] μA :U→[0,1]

[0066] u→μ A (u∈[0,1])

[0067] Among them, u A yes A The membership function, u A (u) is u right A The membership degree of A When (u)=1, u∈ A , when u A When (u) = 0, When u A When (u) is 0 or 1, the membership function degenerates into the characteristic function of the ordinary set. A Will degenerate into a normal set.

[0068] Fuzzy sets are completely characterized by membership functions. In fuzzy mathematics, a number between 0 and 1 is needed to reflect the degree to which an element belongs to a fuzzy set. The membership function is established for this purpose. Whether the membership function can be correctly determined is the key to the effective application of fuzzy set theory. If the fuzzy set is defined in the real number domain, the membership function of the fuzzy set is called a fuzzy distribution. We mainly use the assignment membership function. The so-called assignment method is to apply certain forms of ready-made fuzzy distributions according to the nature of the problem, and then determine the parameters contained in the distribution based on the measured data. The type of fuzzy distribution we use is large-scale. Large-scale fuzzy distribution is suitable for describing fuzzy phenomena that tend to be large, such as "big", "hot", "old", and "thick" colors. The general form of its membership function is

[0069]

[0070] Where a is a constant and f(x) is a non-decreasing function.

[0071] Assume that there are fuzzy subsets on the domain U A1 , A2 ,… A n Constitute a standard model library, if for any element u0∈U, there is

[0072]

[0073] It is considered that u0 relatively belongs to A i .

[0074] The maximum membership principle is a method of fuzzy model identification. Model identification is used to identify which category a specific object belongs to. There are two basic aspects in model identification: a number of standard models are known in advance to form a model library; the object to be identified. In fuzzy comprehensive evaluation, when processing the final result vector and making the final evaluation, the main method used is the maximum membership principle.

[0075] First, the set of psychological risk factors needs to be identified:

[0076] U={U1,U2,U3,U4,U5,U6}

[0077] U1: Extreme psychological tendency U2: Hurting tendency U3: High energy consumption tendency

[0078] U4: Resignation and revenge tendency U5: Negative energy transmission U6: Glass bottle

[0079] U1={U 11 ,U 12 ,U 13 …U 1i}i=[1,14]

[0080] U2={U 21 ,U 22 ,U 23 …U 2j}j=[1,13]

[0081] U3={U 31 ,U 32 ,U 33 …U 3k}k=[1,8]

[0082] U4={U 41 ,U 42 ,U 43 …U 4l}l=[1,12]

[0083] U5={U 51 ,U 52 ,U 53 …U 5m}m=[1,10]

[0084] U6={U 61 ,U 62 ,U 63 …U 6n}n=[1,11]

[0085] U 11 : Depression 12 : Despair 13 : Self-attack...U 1i : Extreme psychological thoughts

[0086] U 21 : Angry U 22 : Hostile U 23 : Impulsive...U 2j : Paranoia

[0087] U 31 : Anxiety 32 :Frustration Tolerance U 33 : Emotional regulation... 3k :pressure

[0088] U 41 : Jealousy U 42 : Revenge U 43 : Interpersonal sensitivity... 4l : Self-centered

[0089] U 51 : Complaint U 52 : Anxiety 53 : Attribution method...U 5l : Depression

[0090] U 61 :VulnerabilityU 62 :Negative life events U 63 : Emotional regulation... 6l : Social support

[0091] Then, it is necessary to determine the weight of each psychological risk profile, that is, to determine the weight set.

[0092] The Delphi method was used to revise the results so that the experts' opinions would be consistent.

[0093] The essence of the Delphi method is to utilize the knowledge, experience, and wisdom of experts to process and gradually correct the information that cannot be quantified and has great ambiguity through consultation.

[0094] The basic steps of the Delphi method are as follows: ① Select experts. Select 10 experts in this professional field who have both practical work experience and deep theoretical knowledge. ② Send the relevant information of the psychological risk indicator system and the unified indicator weight determination rules to the selected experts, and ask them to independently determine the weight values ​​of each group. ③ Processing of the collected results. Sort out and analyze the weights fed back by each expert, and calculate the average and standard deviation of each weight, and require all experts to redefine each weight on a new basis. ④ Repeat ③ until the standard deviation is less than or equal to the pre-given standard ε=0.1, and take the average of each weight estimate this time as the weight result.

[0095] Next, it is necessary to determine the judgment set, that is, the level assessment of psychological risk.

[0096] In order to improve the discrimination of the judgement, this system adopts four levels of judgement, namely:

[0097] V=(V1,V2,V3,V4)

[0098] V1: Poor V2: Average V3: Good V4: Excellent

[0099] After determining the judgment set, the judgment matrix can be generated.

[0100] Suppose for the i-th factor u in the factor set U i Make a judgment on the jth element v in the judgment set V j The degree of membership is r ij , then the i-th factor u i The fuzzy set corresponding to the evaluation result is expressed as:

[0101] R i = {r i1 ,r i2 ,r i3 …r in}

[0102] The membership of each judgment set is combined to form a judgment matrix as follows:

[0103]

[0104] After obtaining the judgment matrix and weight set, a fuzzy transformation is performed to obtain B = A ×R, and then according to the principle of maximum membership, select the score of the item with the highest psychological risk of the test-taker and assess the risk level.

[0105] Compared with the prior art, the present invention has the following beneficial effects:

[0106] 1. The present invention constructs a psychological profile of high-risk individuals by collecting psychological characteristics of extreme psychological people, people who hurt others, and other individuals who have engaged in risky behaviors. According to the fit between the tester and the psychological profile of the high-risk person, the fuzzy comprehensive judgment method is used to distinguish high-risk individuals from non-risk individuals, so as to find the real high-risk population in a more comprehensive, specific, and targeted manner.

[0107] 2. Since the evaluation results of the present invention are for enterprise HR, managers or college teachers to read, we classified them according to the evaluation results of the testers. When the testers are closer to a certain psychological risk profile group, we will focus on describing the characteristics, risk points, and coping measures and management suggestions of this profile group. In this way, even people without basic psychological knowledge can quickly understand the evaluation results, and the suggestions can be more targeted.

[0108] 3. Different from traditional mental health assessment, the present invention not only screens the current psychological problems of individuals, such as depression, anxiety, etc., detects those who currently have psychological problems, and distinguishes them from normal people. In addition, the present invention also screens those individuals who may be on the verge of developing psychological problems due to problems in personality, coping style, psychological resilience and other psychological functions, that is, screens out those individuals who may have mental health problems in the future, and prevents them from happening.

[0109] 4. In addition to testing individual psychological problems, the present invention also tests individual personality, coping style, psychological resilience, defense mechanism, social support and other aspects. This provides a reference for etiology analysis for the clinical treatment of later psychological counselors and psychotherapists, and can more comprehensively and specifically analyze the causes of individual mental health problems.

[0110] 5. Although the evaluation items of the present invention refer to the commonly used evaluation tools at home and abroad, we have not only modified the expression of the items, but also asked psychology experts and non-psychology users to evaluate the comprehensibility and readability of the items to ensure the quality of language and text and reduce the deviations in wording, grammar, and expression habits. In addition, we have built a localized norm for our newly compiled scale instead of directly using foreign norms to ensure the quality of the norms.

[0111] 6. In view of the fact that enterprises or universities are more concerned about what risks and losses their employees or students' mental health problems will bring to the enterprise or school, the present invention not only reports the mental health status of the tester, but also provides the enterprise or school with an explanation of what risks the tester's overall mental status will bring to the enterprise or school. Therefore, the present invention is named a psychological portrait system for intelligently identifying risky personnel based on FMT features.

[0112] 7. Problems that cannot be solved by traditional mathematics can be explained by fuzzy mathematics. Fuzzy mathematics accurately describes the fuzzy extension of things. Comprehensive evaluation is to make an overall evaluation that can reasonably integrate these attributes or factors for things with multiple attributes, or things whose overall quality is affected by multiple factors. Therefore, it is feasible to use the fuzzy comprehensive evaluation method to judge people's psychological risks, and it is also a better way.

[0113] 8. In order to prevent the quality of individual mental health from changing with the change of the reference norm, we have assigned corresponding risk groups to each type of psychological risk, i.e., extreme psychological tendency psychological risk is assigned to people who have attempted extreme psychological behavior, and hurting tendency psychological risk is assigned to people who have hurt others, etc. By setting norms and benchmarking key groups at the same time, we can divide normal people into risky people and different risk level intervals.

[0114] Therefore, the present invention can give a more accurate mental health status. BRIEF DESCRIPTION OF THE DRAWINGS

[0115] Figure 1 A flow chart was prepared for the psychological risk screening scale of the psychological profiling system for intelligently identifying risky personnel based on FMT features.

[0116] Figure 2 This is a flowchart of the fuzzy comprehensive evaluation in the psychological portrait system for intelligently identifying risk personnel based on FMT features.

[0117] Figure 3 This is the UML diagram of the technical system in the system for intelligently identifying the psychological portrait of risk personnel based on FMT features. DETAILED DESCRIPTION

[0118] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0119] See also Figures 1 to 3 In an embodiment of the present invention, a system for intelligently identifying risk personnel psychological portraits based on FMT features includes the following steps:

[0120] S1: Feature screening, key feature collection and risk feature screening;

[0121] S2: Model outline and evaluation of psychological profiling model for high-risk individuals;

[0122] S3: develop the scale, administer the test, and collect data;

[0123] S4: Data analysis, determining the profiling model through Binarylogistic regression;

[0124] S5: Based on the fuzzy comprehensive judgment method, the final risk level of the tester is obtained.

[0125] The key characteristics of S1 were collected by using a semi-structured interview method to interview 10 senior managers from different industries for 30-45 minutes to understand employee risk events that had occurred within the company, such as extreme events such as employee extreme psychology, hurting others, and deliberate sabotage, as well as some risky behaviors that corporate managers were worried about employees implementing. 56 key characteristics of risky employees were obtained. The risk characteristics of S1 were screened by voting to evaluate the key characteristics. Eight senior HRs were invited to vote on each key characteristic, and more than 60% of the items related to the key characteristics of risky employees selected by senior HR were retained. A total of 34 risk characteristics were selected, and 6 types of psychological risk employee character portraits were initially formed, including extreme psychological tendency psychological portraits, hurting tendency psychological portraits, resignation revenge psychological portraits, high energy consumption tendency psychological portraits, negative energy transmission psychological portraits, and glass bottle image psychological portraits.

[0126] The psychological portrait model of S2 medium- and high-risk personnel is mainly drawn by literature analysis. We read and study relevant domestic and foreign literature on different psychological risk portraits, learn and master the risk models of different character portraits. By sorting out and summarizing existing literature and academic achievements, we comprehensively outline 6 types of psychological risk character portraits from 5 dimensions: external stimulation received by individuals, individual emotional state, personality, cognition and psychological resilience. The evaluation of the psychological portrait model of S2 medium- and high-risk personnel includes three levels: employee social function status screening, clinical symptom verification and psychological motivation analysis. Employee social function status screening covers interpersonal quality, behavior efficiency, emotional state, positive feelings and scope of activities. Interpersonal quality assessment includes whether interpersonal relationships are stable and harmonious; efficiency assessment includes the efficiency of memory, reaction, learning and other functions; emotional state assessment includes whether the subject can maintain a stable and peaceful emotional state; positive feeling assessment includes whether the individual has a sense of happiness and accomplishment, etc.; activity scope investigation includes whether the individual has enough rich social interactions and entertainment and leisure activities. Clinical symptom verification mainly analyzes employees' depression, hostility-aggression, paranoia and other psychological problems, so as to clarify the psychological problems through the clinical symptom verification scale. Psychological motivation analysis mainly includes stressful events, defense mechanisms, social support, etc., so as to analyze the psychological motivations of employees.

[0127] The scale was compiled in S3. The initial scale was compiled according to the dimensional composition of different types of psychological portraits. At the same time, the Self-rating Scale for Mental Health Symptoms (SCL-90), the Extreme Psychological Tendency Scale in the Interpersonal Psychological Model, the Barratt Impulsiveness Scale, the Stress Perception Scale (PSS) compiled by Dr. Cohen, etc. were referred to. A total of 450 original scales were compiled. The test and data collection in S3 were carried out in the form of fixed investment, random investment and large-scale investment. Among them, 6 copies were accurately distributed to typical portrait characters in the fixed investment, such as the subscale for measuring extreme psychological tendencies was distributed to testers who had extreme psychological behaviors but attempted extreme psychology, and the subscale for measuring the tendency to hurt others was distributed to testers who had hurt others, etc. Among them, the random investment was randomly distributed to the general normal population. The large-scale investment was mainly placed on the system APP platform for free measurement by users. Through data cleaning and elimination of invalid questionnaires, a total of 16,805 valid questionnaires were collected.

[0128] Data analysis in S4 includes reliability and validity analysis. Reliability is an indicator that reflects the stability and identity of the tester. The "Cronbach's α coefficient" indicator is used to test the stability and identity of the scale. Validity refers to the extent to which a test measures the characteristics it is intended to measure, and is an indicator that reflects the authenticity and accuracy of the measurement. SPSS20.0 is used to conduct exploratory factor analysis on the collected data, and the compiled scale is tested for validity. Inappropriate questions are deleted through factor analysis to obtain good structural validity.

[0129] In S4, the portrait model is determined, and the data of the testers who take different risk type scales are divided into two categories, namely, extreme psychological and non-extreme psychological, hurtful and non-hurting, etc. Binary logistic regression is used to find the independent variables that actually cause individuals to take extreme psychological behaviors, hurtful behaviors, quitting jobs for revenge, etc., and the log-linear model is used for statistical analysis. In the log-linear model, when its binary variables are used as dependent variables, the log-linear regression model becomes a Binarylogistic regression model. The following takes the extreme psychological tendency model as an example to introduce in detail the derivation and solution of the BinaryLogistic model.

[0130] Assume that the nth individual, n∈N, takes extreme psychological behavior with y n Indicates that y n =1 represents that the individual has adopted extreme psychological behavior, otherwise y n = 0, assuming that in theory, there is a continuous response variable y * n Represents y nThe probability of occurrence ranges from negative infinity to positive infinity. If the value of this variable exceeds a critical point m, such as m = 0, the traveler n will take extreme psychological behavior, so:

[0131] When * n >0 time y n =1

[0132] When * n ≦0 when y n =0

[0133] If we assume that the response variable y * n The factors affecting the selection of independent variables, x ni There is a linear relationship between (i=1,2…,m), namely

[0134] y * n =β0+β1x n1 +…+β m x nm +ε j 1.1

[0135] Where: j is the error term, which obeys the Logistic distribution; β0 is the intercept, which is usually called the constant term; β i (i=1,2…,m) is x ni The partial regression coefficient of (i=1,2…,m) is as follows. From formula 1.1, we can know that the conditional probability of the nth individual taking extreme psychological behavior is:

[0136] P(y n =1 / x ni )=P[(β0+β1x n1 +…+β m x nm +ε j )>0]1.2

[0137] By calculating and solving, we can get:

[0138]

[0139] Where P(y n =1 / x ni ) is a nonlinear function composed of explanatory variables (influencing factors), which can be transformed into a linear function.

[0140] First, define the probability of extreme psychological behavior being selected by individual n as:

[0141]

[0142] Then the probability ratio of extreme psychological behavior being selected and not being selected is:

[0143]

[0144] This ratio is called the occurrence ratio of the event, abbreviated as Odds. From formula 1.5, we can see that the larger the p, the larger the odds; the smaller the p, the smaller the odds. In order to measure the degree of influence of a certain independent variable on whether the dependent variable takes extreme psychological behavior, we define OddsRatio, abbreviated as OR, or odds ratio, as follows:

[0145]

[0146] This means that when the influencing factors of other independent variable selection behaviors remain unchanged, the independent variable x ni Change of one unit, the OR corresponding to the dependent variable changes exp(β i ). We used OR to conduct sensitivity analysis on the factors that influence individuals' choice of extreme psychological behaviors.

[0147] Since Binarylogistic regression is a nonlinear model, the maximum likelihood estimation is used to solve the parameters. Assume that there is a population consisting of N individual choice behaviors, Y1, ...YN, from which n cases are randomly selected as samples, and the observed values ​​are labeled y1, ...yn. Let p i =p(y i =1|x i ) is a given influencing factor x i The result y is obtained under the condition i =1, then y i =0 is p(y i =0|x i )=1-p i , so the probability of getting an observation is:

[0148]

[0149] where y i =1 or 0, because the observations are independent of each other, their joint distribution can be expressed as the product of the marginal distributions, as shown below:

[0150]

[0151] Formula 1.8 is called the likelihood function of whether n individuals will have extreme psychology, where:

[0152] In order to find the parameter estimate that can maximize the value of this likelihood function, * Performing logarithmic transformation, we get the following function:

[0153]

[0154] In order to estimate ln(L * ) reaches its maximum value when the population parameter β0…β m , find their partial derivatives respectively, and then set them equal to 0, and get the following formula:

[0155]

[0156] …

[0157]

[0158] As can be seen from the above formula, there are m+1 independent variables in the system of equations, and there are m+1 simultaneous equations to estimate their values. Since all of the above are nonlinear functions, manual calculation and solution are very difficult. This paper uses SPSS20.0 to solve them.

[0159] Through binary logistic regression, we established 6 regression equations to determine whether the tester belongs to a certain type of risk profile.

[0160] Mainly based on the equation derived from the regression, a certain weight is assigned to each measurement dimension. After the tester completes the test, our system background will automatically add up the dimension values ​​related to a certain portrait, and bring the value of each dimension into the equation of different portraits, and calculate the risk value corresponding to different categories of portraits. The closer the risk value is to 100, the more likely it is that the behavior of the person represented by the portrait will be adopted. Taking extreme psychological tendencies as an example, if the tester's extreme psychological tendency risk score is closer to 100, it means that he matches the personality, emotions, behavior and other characteristics of extreme psychological people, and is more likely to adopt extreme psychological behaviors.

[0161] In S5, the fuzzy comprehensive judgment method is used. Since human psychology is often unpredictable, the level of individual psychological risk often involves multiple factors or multiple indicators. At this time, it is required to make a comprehensive evaluation of things based on these multiple factors, rather than evaluating things based on the situation of only one factor. Therefore, the judgment of comprehensive psychological risk mainly adopts the fuzzy comprehensive judgment method. Here, judgment means to evaluate and judge the pros and cons of things according to given conditions; comprehensive means that the judgment conditions include multiple factors or multiple indicators; fuzzy means using fuzzy mathematical calculation methods for processing. Fuzzy comprehensive judgment can also effectively integrate different types of data, including non-normally distributed data. It has stronger adaptability and ultimately obtains an evaluation of the subjects.

[0162] The internal logic of fuzzy comprehensive evaluation:

[0163] Fuzzy comprehensive evaluation is a very effective multi-factor decision-making method that makes a comprehensive evaluation of things affected by multiple factors. Its characteristic is that the evaluation result is not absolutely affirmative or negative, but is expressed as a fuzzy set.

[0164] Membership degree or membership function is the basis of fuzzy mathematics and fuzzy systems. The concept of "membership degree" is introduced to describe the intermediate transition of differences. It is an approximation of accuracy to fuzziness. Its mathematical definition is:

[0165] For a fuzzy set A on the universe U, specify a mapping from U to [0,1]:

[0166] μ A :U→[0,1]

[0167] u→μ A (u∈[0,1])

[0168] Among them, u A yes A The membership function, u A (u) is u right A The membership degree of A When (u)=1, u∈ A , when u A When (u) = 0, When u A When (u) is 0 or 1, the membership function degenerates into the characteristic function of the ordinary set. A Will degenerate into a normal set.

[0169] Fuzzy sets are completely characterized by membership functions. In fuzzy mathematics, a number between 0 and 1 is needed to reflect the degree to which an element belongs to a fuzzy set. The membership function is established for this purpose. Whether the membership function can be correctly determined is the key to the effective application of fuzzy set theory. If the fuzzy set is defined in the real number domain, the membership function of the fuzzy set is called a fuzzy distribution. We mainly use the assignment membership function. The so-called assignment method is to apply certain forms of ready-made fuzzy distributions according to the nature of the problem, and then determine the parameters contained in the distribution based on the measured data. The type of fuzzy distribution we use is large-scale. Large-scale fuzzy distribution is suitable for describing fuzzy phenomena that tend to be large, such as "big", "hot", "old", and "thick" colors. The general form of its membership function is

[0170]

[0171] Where a is a constant and f(x) is a non-decreasing function.

[0172] Assume that there are fuzzy subsets on the domain U A1 , A2 ,… A n Constitute a standard model library, if for any element u0∈U, there is

[0173]

[0174] It is considered that u0 relatively belongs to A i .

[0175] The maximum membership principle is a method of fuzzy model identification. Model identification is used to identify which category a specific object belongs to. There are two basic aspects in model identification: a number of standard models are known in advance to form a model library; the object to be identified. In fuzzy comprehensive evaluation, when processing the final result vector and making the final evaluation, the main method used is the maximum membership principle.

[0176] First, the set of psychological risk factors needs to be identified:

[0177] U={U1,U2,U3,U4,U5,U6}

[0178] U1: Extreme psychological tendency U2: Hurting tendency U3: High energy consumption tendency

[0179] U4: Resignation and revenge tendency U5: Negative energy transmission U6: Glass bottle

[0180] U1={U 11 ,U 12 ,U 13 …U 1i}i=[1,14]

[0181] U2={U 21 ,U 22 ,U 23 …U 2j}j=[1,13]

[0182] U3={U 31 ,U 32 ,U 33 …U 3k}k=[1,8]

[0183] U4={U 41 ,U 42 ,U 43 …U 4l}l=[1,12]

[0184] U5={U 51 ,U 52 ,U 53 …U 5m}m=[1,10]

[0185] U6={U 61 ,U 62 ,U 63 …U 6n}n=[1,11]

[0186] U 11 : Depression 12 : Despair 13 : Self-attack...U 1i : Extreme psychological thoughts

[0187] U 21 : Angry U 22 : Hostile U 23 : Impulsive...U 2j : Paranoia

[0188] U 31 : Anxiety 32 :Frustration Tolerance U 33 : Emotional regulation... 3k :pressure

[0189] U 41 : Jealousy U 42 : Revenge U 43 : Interpersonal sensitivity... 4l : Self-centered

[0190] U 51 : Complaint U 52 : Anxiety 53 : Attribution method...U 5l : Depression

[0191] U 61 :VulnerabilityU 62 :Negative life events U 63 : Emotional regulation... 6l : Social support

[0192] Then, it is necessary to determine the weight of each psychological risk profile, that is, to determine the weight set.

[0193] The Delphi method was used to revise the results so that the experts' opinions would be consistent.

[0194] The essence of the Delphi method is to utilize the knowledge, experience, and wisdom of experts to process and gradually correct the information that cannot be quantified and has great ambiguity through consultation.

[0195] The basic steps of the Delphi method are as follows: ① Select experts. Select 10 experts in this professional field who have both practical work experience and deep theoretical knowledge. ② Send the relevant information of the psychological risk indicator system and the unified indicator weight determination rules to the selected experts, and ask them to independently determine the weight values ​​of each group. ③ Processing of the collected results. Sort out and analyze the weights fed back by each expert, and calculate the average and standard deviation of each weight, and require all experts to redefine each weight on a new basis. ④ Repeat ③ until the standard deviation is less than or equal to the pre-given standard ε=0.1, and take the average of each weight estimate this time as the weight result.

[0196] Next, it is necessary to determine the judgment set, that is, the level assessment of psychological risk.

[0197] In order to improve the discrimination of the judgement, this system adopts four levels of judgement, namely:

[0198] V=(V1,V2,V3,V4)

[0199] V1: Poor V2: Average V3: Good V4: Excellent

[0200] After determining the judgment set, the judgment matrix can be generated.

[0201] Suppose for the i-th factor u in the factor set U i Make a judgment on the jth element v in the judgment set V j The degree of membership is r ij , then the i-th factor u i The fuzzy set corresponding to the evaluation result is expressed as:

[0202] R i = {r i1 ,r i2 ,r i3 …r in}

[0203] The membership of each judgment set is combined to form a judgment matrix as follows:

[0204]

[0205] After obtaining the judgment matrix and weight set, a fuzzy transformation is performed to obtain B = A ×R, and then according to the principle of maximum membership, select the score of the item with the highest psychological risk of the test-taker and assess the risk level.

[0206] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. Based on FMT features, the system for intelligently identifying risk personnel psychological portraits is characterized by: The following steps are involved: S1: Feature screening, key feature collection and risk feature screening; Key characteristics collection: 10 senior managers from different industries were interviewed for 30-45 minutes using semi-structured interviews to understand employee risk events that had occurred within the company. 56 key characteristics of risky employees were obtained. Risk characteristics screening: key characteristics were evaluated by voting. Eight senior HRs were invited to vote on each key characteristic. More than 60% of the items related to the key characteristics of risky employees selected by senior HRs were retained. A total of 34 risk characteristics were selected, and six types of psychological risk character portraits were initially formed: psychological risk portraits of suicidal tendencies, psychological risk portraits of hurtful tendencies, psychological risk portraits of resignation revengers, psychological risk portraits of high energy consumption tendencies, psychological risk portraits of negative energy transmission, and psychological risk portraits of glass bottle images. S2: Outlining and evaluating the psychological profile model of high-risk personnel; The psychological portrait model of high-risk personnel is outlined. Literature analysis method is used to read and study relevant domestic and foreign literature on different psychological risk portraits, learn and master the risk models of different character portraits, and summarize the existing literature and academic achievements. From the five dimensions of external stimulation received by individuals, individual emotional state, personality, cognition and psychological resilience, six types of psychological risk character portraits are outlined. The psychological risk portrait model assessment of high-risk personnel includes three levels: employee social function status screening, clinical symptom verification and psychological motivation analysis. Employee social function status screening covers interpersonal quality, behavior efficiency, emotional state, etc. The assessment of interpersonal quality includes whether interpersonal relationships are stable; the assessment of performance efficiency includes the efficiency of memory, reaction, and learning functions; the assessment of emotional status includes whether the subject can maintain a stable and peaceful emotional state; the assessment of positive feelings includes whether the individual has a sense of accomplishment; the scope of activities includes social interaction and recreational activities; clinical symptom verification includes analyzing employees' depression, hostility-aggression, and paranoia, so as to identify psychological problems through the clinical symptom verification scale; psychological motivation analysis includes stress events, defense mechanisms, and social support, so as to analyze employees' psychological motivations; S3: develop the scale, administer the test, and collect data; S4: Data analysis, the portrait model is determined by the Binarylogistic regression model. The data of the testers who took different risk type scales are divided into two categories, namely, those who have committed suicide and those who have not committed suicide, and those who have hurt others and those who have not hurt others. The binary logistic regression method is used to find out the independent variables that lead individuals to commit suicide and hurt others. The logarithmic linear regression model is used for statistical analysis. In the logarithmic linear regression model, when its binary variable is used as the dependent variable, the logarithmic linear regression model becomes a Binarylogistic regression model; the derivation and solution of the Binarylogistic regression model of the suicidal tendency psychological portrait model are as follows; Assume that the nth individual takes suicide behavior and uses y n Represents, n∈N, y n =1 represents that the individual has committed suicide, otherwise y n = 0, assuming there is a continuous response variable y * n Represents y n The probability of occurrence ranges from negative infinity to positive infinity. If the value of this variable exceeds a critical point m, m = 0, then individual n will commit suicide, so: When * n >0 time y n =1 When * n ≤0 when y n =0; Assume that in the continuous response variable y * n Factors affecting the behavior of independent variable selection x nm There is a linear relationship between y * n =β0+β1x n1 +…+b m x nm +e j 1.1 Where: j is the error term, which obeys the Logistic distribution; β0 is the intercept, which is called the constant term; β m is x nm The partial regression coefficient is: From formula 1.1, we can know that the conditional probability of the nth individual taking suicide behavior is: P(y n =1 / x nm )=P[(β0+β1x n1 +…+b m x nm +e j )>0]1.2 By calculating and solving, we can get: Where P(y n =1 / x nm ) is a nonlinear function consisting of explanatory variables, where the explanatory variables are the influencing factors of the independent variable selection behavior, and the nonlinear function is transformed into a linear function. First, define the probability of suicide behavior being chosen by individual n as: Then the ratio of the probability of suicide behavior being chosen to the probability of suicide behavior not being chosen is: This ratio is called the occurrence ratio of the event, abbreviated as Odds, or advantage. In order to measure the influence of the influencing factors of a certain independent variable's choice behavior on whether the dependent variable takes suicidal behavior, the OddsRatio, abbreviated as OR, is defined as follows: It means that when the influencing factors of other independent variable selection behaviors remain unchanged, the influencing factor x of independent variable selection behavior ni Change of one unit, the OR corresponding to the dependent variable changes exp(β i ), using OR, to conduct sensitivity analysis on the factors affecting individual suicidal behavior, Since the Binarylogistic regression model is a nonlinear model, the maximum likelihood estimation is used to solve the parameters. The overall behavior consisting of n individual choice behaviors is counted as Yn, and i cases are randomly selected as samples, and the observed values ​​are marked as y i , let p i =p(y i =1|x i ) is a given influencing factor x i The result y is obtained under the condition i =1, then y i =0 is p(y i =0|x i )=1-p i , so the probability of getting an observation is: where y i =1 or 0, y i =1 means yes, y i =0 means nothing, because the observations are independent of each other, so the joint distribution of the observations is expressed as the product of the marginal distributions, as shown in the following formula: Formula 1.8 is called the likelihood function of whether n individuals will commit suicide, where: In order to find the parameter estimate that can maximize the value of this likelihood function, * Performing logarithmic transformation, we get the following function: In order to estimate ln(L * ) reaches its maximum value when the population parameters β0, …, β m , find their partial derivatives respectively, and then set them equal to 0, and get the following formula: As can be seen from the above formula, there are m+1 independent variables in the system of equations, and there are m+1 simultaneous equations to estimate their values; S5: The final risk level of the tester is obtained based on the fuzzy comprehensive judgment method. The internal logic of fuzzy comprehensive judgment method: Fuzzy comprehensive judgment method is a multi-factor decision-making method that makes a comprehensive evaluation of things affected by multiple factors. Its characteristic is that it is represented by a fuzzy set; Membership degree or membership function is the basis of fuzzy mathematics and fuzzy systems. The mathematical definition of membership degree is: For a fuzzy set A on the universe U', specify a mapping from U' to [0,1]: Among them, u A yes A The membership function, u A (u) is u right A The membership degree, u A (u) is the mapping of the corresponding set, when u A When (u)=1, u∈ A , when u A When (u) = 0, When u A When (u) is 0 or 1, the membership function degenerates into the characteristic function of the ordinary set. A will degenerate into a normal set, The fuzzy set is completely characterized by the membership function, and the general form of its membership function is Where a is a constant and f(x) is a non-decreasing function, Assume that there are fuzzy subsets on the domain U' A 1, A 2,… A n Constitute a standard model library, if for any element u0∈U', there is It is considered that u0 relatively belongs to A i , The maximum membership principle is a method of fuzzy model recognition. Model recognition is used to identify the category to which a specific object belongs. There are two basic aspects in model recognition: a number of standard models are known in advance to form a model library and the object to be identified; the maximum membership principle is used in the fuzzy comprehensive judgment method when processing the final result vector to make the final evaluation. First, the set of psychological risk factors needs to be identified: U={U1,U2,U3,U4,U5,U6} U1: Suicidal tendency, U2: Hurting tendency, U3: High energy consumption tendency, U4: Resignation and revenge tendency, U5: Negative energy transmission, U6: Glass bottle; U1={U 11 ,U 12 ,U 13 …U 1w }w=[1,14] U2={U 21 ,U 22 ,U 23 …U 2e }e=[1,13] U3={U 31 ,YOU 32 ,YOU 33 …YOU 3k }k=[1,8] U4={U 41 ,And 42 ,And 43 …AND 4l }l=[1,12] U5={U 51 ,U 52 ,U 53 …U 5r }r=[1,10] U6={U 61 ,U 62 ,U 63 …U 6t }t=[1,11] Then, it is necessary to determine the weight of each psychological risk profile, that is, to determine the weight set. The Delphi method was used to correct the results so that the experts' opinions would be consistent. The steps of the Delphi method are as follows: ① Select experts, select 10 experts in this professional field who have both practical work experience and theoretical knowledge, ② Send the relevant information of the psychological risk indicator system and the unified indicator weight determination rules to the selected experts, and independently determine the weight values ​​of each group, ③ Process the collected results, organize and analyze the weights fed back by each expert, and calculate the average and standard deviation of each weight, and require all experts to re-determine each weight on the new basis, ④ Repeat ③ until the standard deviation is less than or equal to the pre-given standard ε=0.1, and take the average of the estimated values ​​of each weight this time as the weight result, Next, we need to determine the judgment set, that is, the psychological risk level assessment. The system adopts four levels of evaluation, namely: V=(V1,V2,V3,V4) V1: Poor V2: Average V3: Good V4: Excellent After determining the judgment set, the judgment matrix is ​​generated. Suppose for the i-th factor u in the factor set U i Make a judgment on the jth element v in the judgment set V j The degree of membership is r ij , then the i-th factor u i The fuzzy set corresponding to the evaluation result is expressed as: R i ＝{r i1 ,r i2 ,r i3 …r in} The membership of each judgment set is combined to form a judgment matrix as follows: After obtaining the judgment matrix and weight set, a fuzzy transformation is performed to obtain B = A ×R, and then according to the principle of maximum membership, select the score of the test-taker’s highest psychological risk item and assess the risk level.

2. The FMT feature-based intelligent identification of risk personnel psychological profiling system according to claim 1 is characterized by: The scales prepared in the S3 are prepared according to the dimensional composition of different types of psychological risk portraits. The initial scales are prepared with reference to the Self-rating Scale for Mental Health Symptoms SCL-90, the Suicidal Tendency Scale in the Interpersonal Psychological Model, the Barratt Impulsiveness Scale, and the Stress Perception Scale PSS compiled by Dr. Cohen. A total of 450 original scales are prepared. The testing and data collection in the S3 are carried out in the form of fixed-shot, random-shot and large-scale delivery. In the fixed-shot, 6 copies are precisely delivered to typical portrait characters, the subscale for measuring suicidal tendencies is delivered to testers who have committed suicide but failed, and the subscale for measuring the tendency to hurt others is delivered to testers who have committed hurtful behaviors. The random delivery is randomly delivered to normal people, and the large-scale delivery is placed on the system APP platform for free measurement by users. Through data cleaning and elimination of invalid questionnaires, a total of 16,805 valid questionnaires were collected.

3. The FMT feature-based intelligent identification of risk personnel psychological profiling system according to claim 1 is characterized by: The data analysis in S4 includes reliability and validity analysis. Reliability is an indicator that reflects the stability and identity of the tester. The "Cronbach's α coefficient" indicator is used to test the stability and identity of the scale. Validity refers to the extent to which a test measures the characteristics it is intended to measure, and is an indicator that reflects the authenticity and accuracy of the measurement. SPSS20.0 is used to perform exploratory factor analysis on the collected data, and the compiled scale is tested for validity. Inappropriate questions are deleted through factor analysis to obtain new structural validity.

Images