CN110600120A - Bayesian theory-based system testing and map analyzing method - Google Patents

Abstract

The invention provides a system testing and map analyzing method based on Bayesian theory, which comprises the following steps: the method adopts a multi-channel instrument for testing, and comprises the following steps: step S1, according to the present evaluation event, the evaluation object is basically tested, if the basic test is passed, the test conclusion is generated, otherwise, the step S2 is executed; step S2, performing fine test on the evaluation object, and generating a test conclusion no matter whether the fine test is passed or not; wherein, the basic test and the fine test both adopt one or more of a quasi-rope problem test method and a hidden information test method.

Description

Bayesian theory-based system testing and map analyzing method

Technical Field

The invention relates to the technical field of data processing, in particular to a system testing and map analyzing method based on Bayesian theory.

Background

The introduction of Bayes Inference (Bayesian Inference) into the analysis of results of multi-channel tests (Polygraph Test) has a long history, and because many psychological textbooks like the use of Bayes theorem in interpretation of lie-detection results, the multi-channel tests and Bayes Inference are very source in view of the "lie-detection" performance of the multi-channel instruments. But even though these interpretations are good and applications are good, it seems always to be a digital interpretation and sometimes very tedious to discuss, but when the use of multichannel instrumentation test results is really needed, the conscious Bayes reasoning seems to disappear, much more so, some doubts and inexplicable questions that are not aware of clouds. The perception is that these two things come together very early, but no healthy baby is born, and instead, either the baby is discarded or the fetus is strange.

The Bayes reasoning succeeds in walking up the court, especially thanks to the advent of DNA technology, whereby the reasoning process starts no surprise, since the "iron evidence" of DNA requires Bayes' theorem to be reasonably explained, to mention nothing else. The authors are also directly inspired by the DNA technology, and only then begin to organize the results of the multi-channel test into a form that meets the forensic evidence requirements by Bayes inference, thereby realizing evidence trust of the first criminal judge in China on the results of the multi-channel test, and thus, completely opening the forensic evidentization channel of the results of the multi-channel test (also called "psychological test").

If the criminal judgment evidentization of the multi-channel instrument test result is applied to the fact evaluation result of Bayes inference, the application of the multi-channel instrument test required for daily evaluation of personnel is the value evaluation result. Among the relationships of factual evaluation to value evaluation, Bayes inference undoubtedly plays a core role. The fact evaluation can be converted into value evaluation, and is also the basis of fundamental change of the test function of the multi-channel instrument, otherwise, the multi-channel instrument is only just a tool.

For a long time, the multi-channel instrument also bears more or less personnel evaluation responsibilities of certain organizations and departments, particularly in the United states, and large-scale multi-channel instrument tester evaluation is started in organizations such as the department of defense and the department of energy after the second world war. However, because such assessment activities lack powerful support of theoretical logic systems, these assessment activities are still controversial, and it is not known why the user is aware of their effects, and the instructor is self-affirmed to be in a reasonable and tricky state.

In fact, from the philosophy level, this is the inevitable result of the unfortunate fall into the "truth-value" and "dichotomy" predicaments. In the multi-channel instrument test of the criminal investigation, people pay more attention to the 'fact' attribute of the test result, and do not have the 'dichotomy', so the dispute is not much. When the multi-channel instrumental tests involve personnel evaluation, the carity appears in fact and value, thereby causing a dichotomy. If the reality is left to "binary" with the value, the transition from "reality" to "value" cannot be realized, and the above-mentioned tricky situation will continue.

The existing multi-channel instrument test pattern analysis only stays on the score, but how to comprehensively judge a large number of pattern scores does not have a corresponding set of mature analysis system. Interpretation of Bayes' reasoning allows one to understand it not only as a mathematical formula, but as a method of understanding, even a world view. This provides a way to achieve the "crossing over" described above.

The system (survey) test (SPEI) is a test method formed by combining specific test practice, systematization and practicability summary on the basis of a psychological information reaction model (two distributions) of a tested person by utilizing an information coupling principle according to a basic idea of information exploration. The system (survey) test consists of a basic test and a fine test, and is a technology which can not only test normal testees, but also effectively test polluted testees and bleached testees through a biased test according to the requirement of minimum test quantity of psychological information.

Multichannel testing techniques and methods do not form a true scientific paradigm and have long been subject to the friye rule in the litigation field.

Compared with the problem of rope alignment, most of the hidden information tests (CIT) are provided, but as pointed out in 2009 by krabbel (Donald j.krapohl) and other people, the design of the hidden information test is only for a tested person who knows specific information; secondly, the target problem is known to specific persons, such as witnesses, victims, and even persons who claim only "said … …", and are not suitable as testees; the confidentiality degree of the target problem has direct influence on the test effect. The ideal choice should therefore be used in conjunction with the quasi-rope problem test, which is a practice for the system survey test (SPEI).

Disclosure of Invention

The object of the present invention is to solve at least one of the technical drawbacks mentioned.

Therefore, the invention aims to provide a system testing and map analyzing method based on the Bayesian theory.

In order to achieve the above object, an embodiment of the present invention provides a system testing and spectrum analyzing method based on bayesian theory, which includes the following steps:

step S1, according to the present evaluation event, the evaluation object is basically tested, if the basic test is passed, the test conclusion is generated, otherwise, the step S2 is executed;

step S2, performing fine test on the evaluation object, and generating a test conclusion no matter whether the fine test is passed or not;

wherein, the basic test and the fine test both adopt one or more of a quasi-rope problem test method and a hidden information test method.

Further, in the basic test of step S1, a combination of two sets of multi-target quasi-rope problem unit tests and one set of hidden information unit tests is set.

Further, the question structure of the quasi-rope question testing method requires that each group of test questions must include three types of test questions, namely, a related question, a quasi-rope question and an unrelated question.

Further, the method for testing the hidden information comprises the following steps: the key question and the accompanying question are used for detecting whether the psychological pressure responses of the evaluation object to the key question and the accompanying question are different or not, and further deducing whether the psychological information of the evaluation object contains the question elements to be investigated or not.

Further, in the multi-channel instrument test, the map probability in the test conclusion is converted into the posterior probability through the Bayesian theorem to realize the test.

Furthermore, in the rope alignment problem test, physiological index data are collected, and various physiological index data are analyzed in a map form.

Further, the physiological index data includes:

(1) respiration takes the breath line length as a measure of Ir and Ic;

(2) skin electrification adopts peak height or peak area;

(3) the blood pressure is raised by the base line, and lowered or increased by the base line.

Further, the implementation of the posterior probability comprises the following steps:

calculate proof weight in multichannel tester WoE:

WoE＝L_{+ Association}/L_{-federation of}；

When WoE or CAI is converted into probability value, the method can be realized by directly adopting the following formula:

therefore, in the Bayesian qualitative assessment,

where WoE is the probability ratio for event B at conditions (A) and (-A).

Further, in the posterior probability implementation, two normal distributions are set according to scores in a regional comparison test by a Bexter scoring technology, and an ROC curve is drawn through a signal detection theory.

Further, analyzing the test pattern, including:

setting physiological index weight, testing pass weight, false negative rate and false positive rate;

selecting a topic mode, and inputting topic contents and topic scores;

and setting output file names and storage path paths of all titles.

The Bayesian theory-based system testing and map analyzing method provided by the embodiment of the invention has the following beneficial effects:

(1) testing method characteristics and functions

The purpose of finding out the accurate psychological information reaction point of the tested person (or suspect) by a test mode of common situation common theory based on information exploration as guidance, psychological information as a core and information coupling can be realized by a well-organized system (survey) test, and an evidence method can be helped to be established. The evidence method is not only beneficial to improving the knowledge of a judicial staff on the multi-channel instrument test, but also can improve the evidence idea of the common people, and can play a larger role on the basis of the original investigation means.

Mainly embodied in the following aspects:

verification of

The verifiability of the fine test is intended to verify that the person under test determined to be related to the case (event) under evaluation by the basic test is such that a level of confidence is achieved. Of course, sometimes the verification can also be embodied in determining whether the tested person has a correlation with the evaluated case (event).

Exploratory property

In the investigation test function, the exploratory test is often used as a detection means, which is helpful for determining the detection direction, finding the evidence, finding the falling of dirt, and the like. In a screening test or an inspection test, exploratory properties often help to further ascertain details of the person under test related to the case (event) under evaluation.

Expansibility

In the criminal investigation test, expansibility means that whether other problems exist besides the fact that a tested person is related to an investigated case or not is solved by using a fine test, the test is set for expanding the fighting results, and a plurality of non-beginners of the same type or a type of criminal case are often tested by using an expansibility test theme in actual combat so as to achieve psychological information required for deeply digging a crime.

Non-uniformity of search

The verification, searchability and extensibility are not achieved for the fine testing of each human being tested. In practice, different types and different quantities of fine test subjects are often set according to the specific situation of a case, and the types and the quantities of the test subjects are flexibly determined and used according to the progress situation of the test.

Systematicness

The basic test and the fine test are not simple and isolated, have inherent connectivity, and need to be carefully set according to the characteristics of people and correspond to each other. The decision as to whether to perform the fine test is made only after the basic test results are obtained, and if it is proved that the person under test "passes" the basic test, the fine test is not performed in general. While when the person under test "fails" the basic test, the fine test must be performed and is usually a spread and drill-in of problems associated with the basic test.

Robustness

Practice shows that the systematic (investigation) test has very good robustness (robustness), i.e. although it is designed for criminal investigation test, its utility is also very suitable for screening test and inspection test, and becomes the solid foundation of multi-channel tester for quality evaluation.

(2) Innovativeness of atlas analysis method

2.1 idea of two distributions

The two distributions provide a base point for scientific judgment of the multichannel instrument test, and the base point becomes a basic reference of the whole system, and the functions of the two distributions are represented as follows: the test method has the advantages that the overall accuracy of the test can be accurately evaluated; an ROC curve is drawn through a signal detection theory, and the overall accuracy of the test can be measured; secondly, providing a basic basis for atlas evaluation and analysis; data support is provided for setting a scoring standard, so that scoring is more rigorous and accurate; the third indicates the direction for the technical improvement; the technical limitation and the advantages are accurately found, and the technical level is improved and improved in a targeted manner.

Depending on the statistics chosen, there should also be multiple types of "two distributions".

2.2 application of Bayesian theorem to dominant expression

In multi-channel testing, bayesian decision making is implemented by the expression of the superiority of bayesian theorem (likelihood ratio L). The advantage expression of Bayes ' theorem not only avoids the influence of the prior probability thoroughly from the mathematical form, but also has the maximum contribution that the joint advantage can be obtained through the joint probability, so as to obtain the evidence weight, and the promotion and the conversion from ' uncertain ' to ' determined ' are completely realized through the integration of all information points. Relative to the accuracy of the past, the odds ratio (or L) is used to judge or is easier to understand and master by using the change of the prior probability and the posterior probability.

2.3 lifting of the Large guide rope

The large guideline is a name for regarding all accompanying lining (background) problems in the hidden information test as the guideline problem, and after the concept is introduced into the multi-channel instrument test, the barrier of data analysis of the guideline problem test and the hidden information test in the traditional sense is broken, and simultaneously, the barrier is cleared for the application of Bayesian theorem in the data analysis of the multi-channel instrument test.

Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.

Drawings

The above and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:

FIG. 1 is a flow chart of a Bayesian theory based system testing and profile analysis method according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of a system test flow according to an embodiment of the present invention;

FIG. 3 is a schematic diagram of two distributions scored as a standard, common in a regional comparison test, by the Bexter scoring technique according to an embodiment of the present invention;

FIG. 4 is a schematic diagram of a electrodermal-3 response according to an embodiment of the invention;

FIG. 5 is a schematic diagram of a electrodermal +3 reaction, according to an embodiment of the invention;

FIG. 6 is a schematic diagram of a electrodermal +2 reaction, according to an embodiment of the invention;

FIG. 7 is a schematic diagram of a electrodermal-1 response according to an embodiment of the invention;

FIG. 8 is a schematic diagram of a galvanic 0 point response according to an embodiment of the present invention;

FIG. 9 is an interface diagram of a test profile analysis according to an embodiment of the present invention.

Detailed Description

Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are illustrative and intended to be illustrative of the invention and are not to be construed as limiting the invention.

The Bayesian theory-based system test and spectrum analysis method of the embodiment of the invention forms the basic components of the test by unit test, and the unit test of the multi-channel instrument test is usually repeated three times. The basic unit of the basic test and the fine test adopts the quasi-rope problem test (CQT) and the hidden information test (CIT) form of the traditional multi-channel tester test. Wherein, the basic test and the fine test both adopt one or more of a quasi-rope problem test method and a hidden information test method.

As shown in fig. 1 and fig. 2, the system testing and mapping method based on bayesian theory according to the embodiment of the present invention includes the following steps:

and step S1, performing basic test on the evaluation object according to the evaluation event, if the basic test is passed, generating a test conclusion, otherwise, executing step S2.

The basic test (BE) is a test designed to satisfy "a test minimum amount requirement capable of forming a test result judgment", and is set for the purpose of determining whether or not there is a relationship between an evaluation target and a commitment event. Especially in the multi-channel meter test of the investigation case, the purpose is to protect innocent, i.e. innocent assessment objects can be provided with 'innocent proof of whitening' through the basic test. If the evaluation object passes the basic test, indicating that it is not associated with the commitment event, the evaluation object is not subjected to the fine test.

In the basic test of step S1, a combination of two sets of multiple multi-target quasi-rope problem unit tests and one set of hidden information unit tests (2+1 combination) is set, which is the backbone form of the basic test.

(1) The 2+1 combination generally comprises no less than 7 related test points (subjects);

(2) any combination of tests that meets the minimum requirements of the test may constitute a base test;

(3) firstly, performing a basic test on a tested person;

(4) the person under test who "passes" the basic test is not required to be subjected to the fine test.

Wherein, the quasi-rope problem test structure is as follows:

the subject structure of the quasi-rope problem testing method requires that each group of testing subjects must include three main types of testing problems, namely, related problems (also called subject problems, denoted by R), quasi-rope problems (also called comparative problems, denoted by C) and unrelated problems (also called neutral problems, denoted by I). Unit testing is typically repeated at least three times. The main structure has two types:

1. single-subject Multi-sided (Multi-faces) structures

The quasi-rope problem test in the system (survey) test is more often used with single-subject multi-sided structures, sometimes also referred to as single-subject multi-target tests, which are often found in basic tests.

The setting of relevant problems in a multi-side quasi-rope problem test unit has certain inherent logicality, the quasi-rope problem is used for triggering basic psychophysiological reaction of a tested person, and no special requirement is made on whether the tested person can lie or not.

Taking a kill protocol test as an example, a typical set of systems (surveys) tests a single subject multi-sided guideline problem test as in table 1. This set of subject matter has the characteristics of a typical R-C-R region comparison structure, with C3 placed at the end, which can be used as a reference for longitudinal comparisons. In addition, as can be seen from the subject content, the set of subjects does not require that the tested person answer 'no' to all questions, so that the pre-test conversation emphasizes the faithful answer of the tested person.

Therefore, the system (survey) test has no provision for rigidity of the format, and can be used as long as the requirement of the title compiling principle is met, thereby providing the maximum flexibility for format selection. In the following, several quasi-rope problem unit test subject formatting formats are recommended for reference, still mainly in the form of single-subject multi-face (multiple-face):

i1 (irrelevant) -Sr 1 (victim relevant) -C1 (quasi-roping problem 1) -R1 (relevant problem 1) -C2 (quasi-roping problem 2) -R2 (relevant problem 2) -C3 (quasi-roping problem 3) -R3 (relevant problem 3).

②I1—I2—I3(Sr)—R1—C1—R2—I4—R3—C2—R4—I5—I6

③I1—I2—R1—C1—R2—I3—R3—C2—R4—I4

④I1—I2—C1—R1—I3—C2—R2—I4—C3—R3—I5

⑤I1—I2—R1—C1—I3—R2—C2—I4—R3—C3—I5

⑥I1—I2(Sr)—R1—C1—R2—I3—R3—C2—R4—C3—I4

2. Single-theme You-only (Single-issue Phase) structure

Corresponding to the single-subject multi-sided test format is what is commonly referred to as a traditional single-subject one-only test, which is not frequently used in system (survey) tests, and is sometimes used in fine tests as an addition to covert information tests. Table 1 shows the subject content and structure of a single subject, you only test for a set of system (survey) tests.

This group of items 3, 4, 5 constitutes an SKY test structure, which is equivalent to the test structure form of two quasi-ropes and one primary (main) related problem.

TABLE 1 System (survey) test Single subject you Only test

The hidden information test structure is as follows: the hidden Information Test (CIT) is a generic name of a class of psychological Test methods, and the core of the method is composed of Key questions (or related questions, theme questions, target questions, etc., denoted by K) and accompanying questions (or Background questions, denoted by B). The CIT is mainly used for detecting whether psychological stress responses of the assessment object to the key questions and the accompanying questions are different or not, and further deducing whether the psychological information of the assessment object contains the question elements to be investigated or not. The evaluation of the reauthorization capability of the evaluation object is the key for the necessary assurance of CIT detection.

The hidden information tests all belong to single-subject single-target tests.

General form:

first (I1) is not related to problem one; (I2) irrelevant problem two; ③ (B1) background problem one; fourthly (B2) a second background problem; fifthly, (K) key problem; sixthly, (B3) problem of background three; seventhly (B4) solving a background problem IV; problem three of no relevance (I3).

Problems (c) to (c) are generally to disturb the sequence after the problem (c) and continue the 2 nd test, and disturb the sequence again and conduct the 3 rd test.

In the present invention, the stress peak test (POT) and the crime episode test (GKT) are special cases of the hidden information test, and are not generally distinguished particularly in the system (survey) test.

The known (hidden) information test is a test method for a specific crime episode and is used for detecting whether a tested person knows the specific crime episode or not so as to judge the relevance degree of the tested person and a specific event. The test with known (covert) information must satisfy the following conditions:

firstly, the security state of the criminal plot (information) is good and is not disclosed, and other irrelevant personnel can not know the criminal plot (information);

② the criminal scenario (information) must be accurate and reliable.

Third, there is reason to think that the criminal plot (information) is the conscious action result of the criminal or can have profound influence on the criminal's memory.

Fourthly, the tested person is definitely unknown.

The unknown (hidden) information test is a test method for detecting whether the tested person knows the relevant question or not by utilizing the psychophysiological reaction intensity of the tested person to the relevant question, and is also called as a search test or a detection test.

For the evaluated case (affair), when the situation that only the person of the party or the agent knows needs to be explored, the unknown (hidden) information test can be used for testing the person to be tested;

secondly, the key problem which is not determined in the test subject of the unknown (hidden) information test must include all the possible problems which can be considered, namely, the problem(s) to which the tested person is more relevant are detected through the subject test, so as to probe the unknown information in the evaluation case (event).

The Big quasi-rope problem Test (BCQT) is actually a hidden information Test, and is a general name of a multi-channel instrument Test method, and the core problems include two types: related problems are called as key problems (K) in the hidden information test; the second is the problem of large alignment, which is called as accompanying problem, background problem (B) in the hidden information test, and there are several problems, and many problems are similar and same attribute with the related (key) problems.

And step S2, performing fine test on the evaluation object, and generating a test conclusion whether the fine test is passed or not.

In particular, the fine test (FE) is used to detect a human subject determined to require further detection by the basic test, i.e., a human subject whose basic test determines that it has a correlation with the case (event) under evaluation, and thus is an aid and extension of the basic test. Mainly used for solving the problems of how to correlate and how deeply the tested person who does not pass the basic test and the evaluated case (event). To achieve this, the fine testing is required to achieve the requirements of verifiability, searchability, and extensibility. In the multi-channel tester, the test is generally composed of a series of (n) CIT tests, and n is preferably more than or equal to 3.

The invention has the following test requirements on a multi-channel instrument:

1. properly distinguishing base tests from fine tests

The purpose of the basic test is to protect innocent and thus in principle should cause the least disturbance of the test to innocent, which is also a valuable embodiment of the principle of testing the minimum amount. The fine test is a test performed on the tested talents who are 'failed' in the basic test, so that the fine test has a function of identifying guilty in a certain sense. Therefore, as for the system (survey) test, the basic test is guaranteed, is the central content of the test, and needs to be implemented with the best quality guarantee; the fine test is a mining type, which is an extension content of the test, and the depth and the extent of the extension and the stretching can be determined according to situations. It is obvious that these two methods have reasonable meaning.

2. Test inner question structure of reasonable configuration unit

The proper bedding is intended to allow a smooth transition of adaptation to the subject under constant challenge, especially between the start of the test and the critical problem group. The test structure uses a plurality of test subjects, the stimulus intensities of the test subjects are different, and the reasonable configuration of the problem of different stimulus intensities is very important. Relatively speaking, the strong stimulation test subject has high stimulation intensity on the evaluation object, the evaluation object is relatively tense, waveform disorder is caused, and the evaluation object cannot be subsided for a long time, so that subsequent subject problems are difficult to analyze. Therefore, placing a weak stimulus test theme in front of a strong stimulus test theme is a suitable system configuration.

3. Rational configuration of unit test room structure content

System (survey) test structures are test (stimulus) systems consisting of unit tests, and how to combine the test structure contents is not an arbitrary process. The reasonable distinction between the basic test and the fine test is the first step of the structure configuration, and then the presented stimuli are classified according to topics under the framework of the basic test and the fine test, and the similar topics are concentrated in certain unit test(s) as much as possible, so that the unit tests have independence and contact. Some units in the crime investigation test ask questions aiming at material evidences, some units ask questions aiming at the position of a tested person, and the like, and the general test subject problems are arranged according to the whole test thought system, so that the tested person can keep continuous and stable attention in the whole test process, each unit has a clear test subject, the tested person can easily identify and concentrate attention, and great convenience can be brought to atlas evaluation and judgment.

The multi-channel instrument testing technology is a branch of experimental psychology, and data acquisition of each multi-channel instrument test can be said to be a control experiment in the aspect of experimental psychology, so that the whole process of the experiment must be effectively controlled by data acquisition personnel. If the evaluator is unable to effectively control the overall process of data acquisition, the data acquisition is likely to fail, thereby resulting in failure of the entire multi-channel test. Whether an evaluator can effectively control the data acquisition process of various multi-channel instrument tests or not and whether various evaluation objects of different types can be effectively controlled and effectively matched with the whole test process directly reflects the test professional quality and level of the evaluator.

The control in the acquisition of test data of a multi-channel tester mainly comprises the following three aspects: firstly, controlling a test device; controlling the testing process; and controlling the evaluation object. The most important of which is the control of the evaluation object.

The control of the test equipment means that an evaluator gives full play to each function of the multi-channel tester, the multi-channel tester is flexibly used, the efficiency of the multi-channel tester is given full play, and the multi-channel tester can be timely found and correctly treated when the multi-channel tester breaks down or is abnormal.

The control of the test process refers to that an evaluator correctly adjusts the test rhythm according to the actual condition of data acquisition, so that relaxation is achieved, emergencies occurring in the test process are treated at ease, and the test process is closely matched with the rhythm of the evaluator. Control of the testing process requires the evaluator to be familiar with the details of the events involved in the test, with the testing protocol, and with some ability to review the chart immediately. Namely, the ability of the time evaluation chart is a necessary condition for flexibly controlling the test process and adjusting the test rhythm. If necessary, the evaluator can properly adjust the original test scheme to obtain the best test effect.

The control of the evaluation object means that the evaluation person controls the emotion and state of the evaluation object to the degree that is most suitable for the data acquisition requirement by using a normative language, a proper attitude, a strict program and a conversation in place. The control of the evaluation object is the most important loop in the test control, and the normal data acquisition can be ensured only by effectively controlling the evaluation object really. The most effective and most common method of controlling an object under evaluation is context construction, which when properly applied will have unexpected results.

The test emphasizes the map evaluation and analysis principle of 'corresponding to the question map and giving priority to the question', whether the setting of the test question is reasonable or not is an important premise for forming an ideal test map, the setting of the question is unreasonable, and the beautiful test map also has no analysis value. In addition, the analysis and judgment of the atlas by the system (survey) test not only requires microscopic judgment on each problem, but also requires macroscopic analysis on the action and effect of each problem in the whole system. In a system (investigation) test, Bayesian decision is realized by converting map probability into posterior probability through dominant expression of Bayesian theorem. Namely, in the multi-channel instrument test, the graph probability in the test conclusion is converted into the posterior probability through the Bayesian theorem.

(1) Prior probability acquisition

1. Raw score generation

1) Score standard of seven-point system

Due to the introduction of the seven-point scoring method, the multi-channel tester has great technical capability and is one step higher.

Each "correlation-comparison problem" pair, called a score point (spot);

a seventh grading system for grading the evaluation target into 7 grades according to the difference degree of the reaction intensity of the evaluation target on the grading point; a negative (-) value is assigned when the associated problem response intensity is greater than the comparative problem response intensity, and a positive (+) value is assigned otherwise.

The responses of the "correlation-comparison problem pairs" are compared if:

if there is a little difference, get + -1 point,

② when the difference is obvious, get + -2 points;

thirdly, when the difference is large, plus or minus 3 points are obtained;

and fourthly, when no difference exists, 0 point is obtained.

The number assigned by assigning the atlas data using the 7-point principle described above with reference to the example is called the original score (x), and is typically only a positive or negative integer or zero.

In the posterior probability implementation, two normal distributions are set according to the scores of the Bexter scoring technology in a regional comparison test, and an ROC curve is drawn through a signal detection theory.

According to the method of the signal detection theory, in the multi-channel instrument test, according to the design idea of the quasi-rope problem test, the quasi-rope and the correlation can be mutually signals and noises. In the course of the qualitative assessment, for a clear assessment subject, the guideline problem stimulus is a signal and the associated problem stimulus is noise, and for a "problem" assessment subject the opposite is true. Therefore, in a multi-channel test, "two distributions" can be formed.

The two distributions scored as a standard, one spread around +6 or +9 and the other spread around some-6 or-9, which are common in the Zone Comparison Test (ZCT) according to the Bexter scoring technique. The four-region comparison test of Matte (Matte) gave two distributions: + 6.0. + -. 3.1 and-4.1. + -. 2.8.

As shown in FIG. 3, where one normal distribution on the left is referred to as a "through" domain distribution: -0.33 ± 0.17% Rr; while one normal distribution on the right is referred to as a "no pass" domain distribution: rr is 0.50 ± 0.25.

Here Rr is called the correlation coefficient, and: rr ═ Ir-Ic/Max (Ir, Ic), where Ir and Ic represent the reaction strengths of the related and roping problems, respectively. According to the experiments and studies of the authors,

when Rr is more than or equal to 0.5, namely Ir is more than or equal to 2Ic, x is equal to-3;

② when Rr is more than 0.5 and is more than or equal to 0.3, x is-2;

③ when Rr is more than 0.3 and is more than or equal to 0.15, x is-1;

when Rr is less than or equal to-0.33, namely Ic is more than or equal to 1.5Ir, x is 3;

when-0.25 is not less than Rr > -0.33, x is 2;

sixthly, when Rr is more than or equal to-0.15 and is more than or equal to-0.25, x is 1;

when 0.15 is more than or equal to Rr > -0.15, x is 0.

In the rope alignment problem test, physiological index data are collected, and various physiological index data are analyzed in a map form.

In an embodiment of the invention, the physiological indicator data comprises:

(1) respiration takes the breath line length as a measure of Ir and Ic;

(2) skin electrification adopts peak height or peak area;

(3) the blood pressure is raised by the base line, and lowered or increased by the base line.

Specifically, the conventional alignment problem test refers to the case where the alignment problem and the related problems belong to 1:1, that is, an atlas evaluation mode in which the related problem is compared with only one alignment problem in the nearest vicinity, and is also the embodiment of the bezier area comparison (ZCT).

In the analysis of the conventional guideline problem test spectrum, the reaction intensity Ir and Ic is determined according to different physiological indexes, the reaction characteristics are different, the length of a breathing line is taken as the measurement of Ir and Ic in respiration, the peak height or peak area is taken as skin electricity, and the rise (or fall) or amplitude of blood pressure is taken as a base line. It should be noted that consistency of intensity units (cells) must be maintained in one pass of the test pattern analysis. The following explains the map analysis of each index.

Breathing

The change of the breathing pattern can be easily observed by naked eyes, but compared with the change of the skin electricity and the electrocardio (including blood pressure, pulse, blood volume and the like), the change of the breathing pattern is easier to be consciously controlled by a tested person, so the change of the breathing pattern is also an important index for judging whether the tested person carries out a counter test or not.

The respiratory map evaluation window is related to the number of test passes, generally the evaluation window of the 1 st test is opened when the tested person responds, and is closed after at least 3 respiratory cycles. The evaluation window throughout the 2 nd test may be opened from the end of the tester's question. Note that the relative question needs to be consistent with the length of the evaluation window of the guideline question.

Although the assignment of the respiratory index can be accurately realized through the length of the respiratory line, people in practice can also use the frequency of respiratory characteristic change to perform empirical assignment, and experiments show that the difference of the two obtained scores does not generally have subversive (significant) influence on the test result. The respiratory characteristics changes and assignments are shown in tables 2 and 3 below.

TABLE 2 breathing characteristic Change

TABLE 3 breath assignment

Note that: the absolute value of the respiration index score is less than 1.

② skin electricity

The skin electricity is the most sensitive parameter in multi-channel instrument test, the research on the skin electricity is always the focus of multi-channel instrument test technology, the central position of the atlas analysis has not been shocked since the appearance of the skin electricity, and the value is obvious.

The skin electricity index can be assigned according to the peak height change or the peak area change, and the comparison units are kept consistent in the Rr calculation.

Examples of galvanic skin responses are as follows:

as shown in fig. 4, the electrodermal-3 response (guideline front, correlation back), Ic-3 intensity units, Ir-7 intensity units,

rr ═ 0.57 > 0.5, (7-3)/7 ═ 3 points.

As shown in fig. 5, the galvanic skin +3 response (guideline front, correlation back), Ic-3 intensity units, Ir-2 intensity units,

rr ═ 0.33 (2-3)/3, giving a score of + 3.

As shown in fig. 6, the electrodermal +2 response (guideline front, correlation back), Ic 2 intensity units, Ir 1.5 intensity units,

rr (1.5-2)/2-0.25 > -0.33 to give a +2 point.

As shown in fig. 7, the electrodermal-1 response (guideline front, correlation back), Ic 1 intensity unit, Ir 1.2 intensity unit,

rr ═ 0.16 > 0 (1.2-1)/1.2, giving a score of-1.

As shown in fig. 8, the skin charge is 0 point response (quasi-cord before and after), Ic is 3 intensity units, Ir is 3 intensity units, and Rr is 0, which gives 0 points.

③ electrocardiac reaction

The electrocardio-reaction is commonly the blood pressure and pulse rate change, and the detection modes such as blood volume change also exist at present, but in the test of the prior multi-channel instrument, compared with the respiration and skin electricity change, the electrocardio-change has the self effect greatly influenced by the limitation of the wearing mode of the sensor (the sensor can not be worn like the electrocardio monitoring in a hospital), so the electrocardio-index is generally given smaller weight.

However, when the breath and the skin electrogram reaction of the tested person are excessively slow or obvious due to some reason, the electrocardio reaction can exert the unique effect.

The electrocardio index empirical assigning standard and process can refer to the processing mode of a respiratory map, and in addition, the situation that the absolute value of the electrocardio index assigning is more than 1 rarely occurs.

The large quasi-rope problem test (BCQT) assigned standard is as follows: the large alignment is a name for regarding all accompanying lining (background) problems in the hidden information test as the alignment problem, and after the idea is introduced into the multi-channel instrument test through Chen Yunling, the barrier of data analysis of the alignment problem test and the hidden information test in the traditional sense is broken, and meanwhile, the barrier is cleared away for the application of Bayesian theorem in the multi-channel instrument test data analysis.

A7-point rule can be used for setting a point rule for the traditional alignment problem test, and the rule is also applicable when a concept of a large alignment problem test is introduced, namely when a hidden information test is marked.

In scoring for the large raglan problem test, Ic should be the geometric mean of the reaction intensity of the cossette problem. That is, if the informational test is concealed for a known result, Ic is only generated by the compassing problem; if the informational test is masked for unknown results, Ic is generated by all relevant + companionic questions, then each question is calculated with this Ic to derive an Rr value, and a score is generated based on the criteria previously described.

The choice of Ic as the geometric mean of the reaction intensities of the companionship problem is based on the rationale of Stevenson' Law, one of the psychophysical laws.

In practice, empirical assignment can be adopted, which is exemplified by a known result hidden information test, and when the reaction strengths of the target problems are ranked as follows:

(1) when the intensity of the light source is in the first intensity reaction and the second intensity reaction, obtaining a negative score;

(2) when the intensity reaction is carried out in the third and later intensity reactions, 0 point is obtained;

(3) when the intensity difference between the first intensity and the second intensity exceeds 50 percent, obtaining a score of-3;

(4) when the intensity difference between the first intensity and the second intensity is 30-50%, obtaining-2 points;

(5) when the intensity difference between the first intensity and the second intensity is 15-30%, obtaining a score of-1;

(6) when the intensity difference between the first intensity and the second intensity is 0-15%, 0 point is obtained;

(7) when the intensity difference between the second intensity and the third intensity exceeds 50 percent, obtaining a score of-2;

(8) when the intensity difference between the second intensity and the third intensity is 20-50%, obtaining a score of-1;

(9) when the intensity difference between the second intensity and the third intensity is 0-20%, 0 score is obtained;

(10) the big guide rope (CIT) does not give positive score.

The weighted score (λ) refers to that the original score of the assigned point can be converted into a weighted score through the weighting processing of the original score (x) so as to perform corresponding conditional probability conversion. In the multi-channel instrument test, the weight (weight) is mainly embodied in two aspects, namely the physiological parameter weight (ki) and the test pass number weight (qj). The weighted score (lambda) can be calculated according to the corresponding weight distribution.

Expressed by a matrix equation as follows:

in the formula, xij is the original score of seven scores, ki is the weight of the physiological index, and qj is the weight of the test pass.

Probability acquisition is according to psychological information distribution rule, combines corresponding experimental data, when certain dividing point weighting score is lambda, has promptly:

P(+|D)＝F1(λ)，λ＜0

P(–|T)＝F2(λ)，λ≥0

wherein:

p (+ | D) ═ P positive | spoofing

P (- | T) ═ P negative | honesty

The two are atlas probabilities, which can be obtained by scoring the atlas, and have the attribute of prior probability in Bayesian theorem, which can also be called conditional probability. Wherein, P (+ | D) refers to the occurrence probability that the atlas reaction is positive (negative score, non-passing) when the tested person carries out cheating; p (- | T) refers to the occurrence probability that the tested person shows honesty real time and the map reaction is negative (positive score is obtained, and the map reaction is passed).

In the practice of quality assessment, the following empirical formula has been summarized and extracted for reference:

when λ < 0, P (+ | D) ═ 50.00-5 λ 3/6+0.18 λ 2-9.36 λ (I)

When lambda is more than or equal to 0, P (-T) 5 lambda 3/6-0.18 lambda 2+9.36 lambda +52.14 (II)

At the same time, it is specified that:

(1) when lambda is 3 min, take P (+ | D) ═ 99.9%

(2) When the lambda is-3 min, the P (- | T) is 99.9%

The probability conversion in the multi-channel instrument test data analysis is a process of converting the graph spectrum probability P (+ | D) into P (D | +) by using Bayesian theorem, namely, the probability that deception is to be implemented for a tested person with a positive result is converted into the probability that deception is implemented for the tested person with a positive result. In the same way, P (- | T) is converted into P (T | -), namely the probability that the result of the honest tested person is negative is converted into the probability of the honest person in the tested person with the negative result.

(2) Acquisition of dominant expression

1) Assigning points (spot)

Scoring of spectra (Score) is an initiative and a great progress in the development process of multi-channel testing technology, and is a key point that Skill (Skill) can become technology (Technique).

Isolated responses are often meaningless, psychophysiological responses can only be reflected by comparison, and scoring of multi-channel atlas responses is a direct reflection of such comparison. One point of differentiation contains at least two stimulation questions, namely 1 correlation question and 1 guideline question.

The alignment problem test design in a multi-channel instrument is a standard 1:1 dividing point.

The hidden information test is divided into two types, one is a comparison between a related problem and a plurality of quasi-rope problems (comparison problems and accompanying lining problems), wherein the related problem is a Target problem and is known or known (i.e. wanted to be known), and is generally called a known result hidden information test, and the situation can be called as a 1: n division point, namely a Target (related) problem and a plurality of accompanying lining (quasi-rope) problems (n); the other is that the target is unknown, but one or several targets are to be detected by the hidden information test, that is, in a group of seemingly unrelated problems, the contents which are probably more interesting and interesting to the tested person are searched out, and the condition is called (q: n) dividing point, which is called the hidden information test of unknown result or the hidden information test of searching (searching).

For example, if money in a safe is stolen, the test for the stolen money can form a concealed information test under two conditions of 1: n and q: n.

First case 1: n

This is used to test what the stolen item is, known as money, and what is detected is the reaction of the person under test to the target problem, money. (money and money can be treated separately, but still known)

Situation of q: n

For the destination of the test money, the detection personnel and the testing personnel are unknown to the direction (q), and the destination of the money is not necessarily a direction (q), the range can be defined by setting n subjects, but q problems possibly occur to react in the actual test, so that the direction is marked as q: n.

Strictly speaking, q: n or q: n should not be called a point of division, but for descriptive convenience, the point of division is still used for labeling without ambiguity.

The assigning needs to directly reflect the characteristics of the assigning points, which is the basis and foundation for assigning.

1:1 assigning points:

only 1 point of assignment of the correlation problem and 1 point of assignment of the comparison problem, which basically constitutes the Zone Comparison Test (ZCT) of bexter, is characterized by typical symmetry, and needs the basic support of "two distributions", and theoretically, the "two distributions" should be completely symmetrical. The "two distributions" of the system (survey) test mental information is one representation of such a distribution.

The assigning requirements are as follows: the seven divisions (including five divisions, three divisions, etc.) are apparently designed for the (1:1) division point. It is also characterized by symmetry. With symmetric distribution, the assigned score threshold can be determined, and the threshold is different according to different scores, but the basic idea is consistent, namely, the overall uncertainty degree is explained by score depiction.

The zero point (0) is centered due to the emphasis on symmetry, and the symmetric equidistant expansion of the positive and negative points is the basic characteristic of the division. The feature of the equal distance determines that the scores can be added and subtracted, namely, only the mode of arithmetic mean can be adopted in the process of averaging. The aforementioned weighting of the raw scores is also an extension of the arithmetic mean.

And (3) calculating the advantages of the assignment:

when a related problem (R) of a multi-tester test is compared with a comparative problem (C), which is common in the quasi-roping problem test, based on the assumption of the equality of the quasi-roping problem and related problems to innocent and guilty, it can be considered before the test that:

PD＝PT＝0.5

(ii) Positive results

Referring to the situation that the reaction strength of the relevant problem is greater than that of the quasi-rope problem, by using advantage calculation, the combined formula (VII) has the following steps:

in the formula, R + represents that the reaction intensity of the related problem is greater than that of the quasi-rope problem, namely a positive reaction occurs; p_(+|T)Is to use the prior probability of the quasi-roping problem and related problems, which can be assumed to be equal due to the equality of the quasi-roping problem and related problems_(+|T)＝0.5。

Then the process of the first step is carried out,

more generally, P is_(+|T)If > 0.5, then assuming a difference from 0.5, then the deceptive advantage for a positive reaction would be raised as:

negative result-

The case that the reaction strength of the relevant problem is smaller than that of the quasi-rope problem is indicated, and the improvement of honest advantage of negative reaction can also be obtained:

in which beta is P_(-|D)It is commonly referred to as the false negative rate.

③ general formula

For a related problem and a comparative problem comparison test, the advantage promotion range is as follows:

2)1: n division point characteristics

The resulting covert information test is known to be a typical representation of such points of demarcation. Its characteristics represent at a glance a "one-flower-one-show" which is an isolated distribution of related (target) problems, and a "one-group-male-group-together" which is a numerous distribution of companionship (comparative) problems. This seemingly obvious difference from the (1:1) differentiation point not only results in a long-term discrepancy between the guideline problem test and the hidden information test pattern analysis, but also causes confusion and even a contradictory state for interpretation of the test mechanism.

The assigning requirements of the 1: n assigning points are as follows: the known result hidden information test has the characteristics of' one flower and one flower but the main purpose is to clearly show the inherent connection with the quasi-rope problem test when the meaning of the relevant reaction is obtained by comparison. The large alignment problem presented by system survey and test is the disclosure of the relationship.

The strength characteristics of the problem are not considered too much because the target problem is only one, but the companionship problem "one group" can be effectively compared only by comprehensively processing the problem into "one", namely, taking some representative value (generally, a mean value) of the problem. The Stevens' law of psychophysics gives a solid foundation for this integration.

According to Stevens's law, the mean of the reaction intensities of all companionship problems should be the geometric mean of their respective reaction intensities. Thus, in a one-to-one comparison alone, it is possible to provide for a (1:1) assignment pattern using the guideline problem test, provided that the geometric mean of the companionship problem scores in the known result covert information test can be converted to an arithmetic mean.

The logarithm of the geometric mean is an arithmetic mean of the logarithms of the variable values, that is, the arithmetic mean can be performed by taking the logarithm of the geometric mean. Although the values of the variable values change after being logarithmized, the overall change is consistent, namely, the basic requirement of scoring can be realized.

However, the logarithm definition requirement is that negative numbers and zero numbers have no logarithm in a real number definition domain, so that the scoring of the large quasi-rope problem test can only adopt a one-way mode, namely, in order to be consistent with the scoring standard of the quasi-rope problem test in practical operation, the hidden information test is not assigned with a positive score, and the zero score is taken as a mathematical extreme value.

These two special cases, just as useful for the negative (reaction-non-specific) result description of covert information tests.

1, n assigning points: the predetermined prior probabilities for positive results (reaction "specificity") and negative results (reaction "no specificity") are not the same and need to be processed separately.

In addition, the relevant (target) problems of the covert information test are divided into two cases of known results and unknown results, and here, although the test of the known results is mainly aimed at, the test of the unknown results needs to be converted into the test of the known results to be processed, so that the two tests are different in practical operation, and the special points of the two tests are explained in the discussion of the points of assignment.

(ii) Positive results

For a given result (1: n) point, since the correlation (target) problem is clear and the number of comparison (background) problems is determined, when 1 target problem (correlation problem prepared for "spoofing") and n comparison problems are selected, there are n +1 correlation problems (including target and comparison) in total. Therefore, the false positive rate of the evaluation target is 1/(n +1), and thus it can be considered that:

and

the target problem is not symmetrical as in the case of (1:1) division point, as compared with the multiple lining (guideline) problem, and therefore needs to be dealt with separately.

A. First positive results

At this time, the known result is consistent with the target problem response and is in the top position, namely the target problem response intensity is greater than the companionship problem response intensity, the above assumptions are substituted into a Bayesian equation, and the result is obtained by combining the dominance calculation:

L_(1:n)+＝(n+1)P_(+|D)

if false positives are also to be considered, i.e. more generally some:

wherein alpha is the false positive rate.

B. Secondary (second) positive results

That is, the reaction intensity of the known target problem is smaller than that of one companionship problem and larger than that of other companionship problems, and the advantages of the reaction are obtained by calculating and combining the advantages.

The reason why the positive result is called positive is that the positive result is less than one accompanying problem reaction and greater than other reaction strengths, so that the positive result can be regarded as one accompanying problem in the case of the first case, namely, (n +1) is changed into n, and in this case:

L+＝(n+1-1)P_(+|D)＝nP_(+|D)

however, since there is a cossette problem with a greater reaction intensity than the target problem, there is now a "negative" reaction to the target problem, i.e. in addition to the "positive" gain of advantage, the "negative" gain of advantage is also taken into account, since there is only one cossette problem, which can be treated as a 1:1 division point (the traditional guideline problem), namely:

L–＝2P_(-|T)

the influence of false positives and false negatives is not considered at this time. Then there are:

clearly, given the same value of P, the second targeted problematic response is less than the first targeted response. This is a basic basis for assigning (-3) the target problem response that does not lie secondary to the hidden information test score described above.

In practical applications, as long as the reaction score of the target problem is negative, it can be confirmed that the reaction intensity of the target problem is greater than the geometric mean value of the reaction intensities of the companying problems, and then the formula (ix) can be directly applied to obtain the relevant superiority value.

Negative result-

Here, the negative result actually includes the case of B in fig. (positive result in the next place (second place)), that is, in a broad sense, when the problem reaction of interest for which the result is known is not in the first place, that is, a negative result reaction occurs.

Substituting the prior probability hypothesis in (1) into Bayesian equation for P_(-|T)And combined with the advantage calculation to obtain

It is clear that this is the advantage obtained for a compassing problem when the target problem "negative (no specificity)" response is made.

When the reaction strength of the target problem is smaller than that of the companding problem, then:

L(1:n)-all＝n[(1/n)(n+1)]P_(-∣T)＝(n+1)P_(-∣T)

therefore, when calculating the "negative (no specificity)" reaction advantage of the target problem, it is necessary to first figure out the number of the accompanying problems to be addressed.

Let r be the number of chaperone problems for the "negative (no specificity)" reaction of the target problem, apparently only in"negative (no specificity)" of the target problem is significant, otherwise the target problem may appear as a positive (specific) reaction. Therefore, the minimum advantage of a known result of the (1: n) score point for use as "honest" is obtained as:

if the effect of false negative β is also to be considered at this point, then:

n points of the score:

unknown results covert information tests are a typical representation of such points of demarcation. Compared with the known result hidden information test in the investigation test, the hidden information test of the quality evaluation (such as the entry evaluation and the on-duty evaluation) is a typical search hidden information test aiming at the search of unknown results. The character expression is that only 'the male and the female are combined' at a glance, and 'a flower and a single show' is not available yet. However, the evaluation effect can not fully show the significance and value, and the delicacy and profound effect under the condition.

Assigning request of q: n assigning points

It is the unique efficacy of this test technique that the probing (searching) hidden information test cannot distinguish the related problem from the collimated rope problem before testing, but distinguishes them by testing. Also, because of this, it can often be used as a miracle effect and even as a performance item for multi-channel testing in research evaluation.

In the quality evaluation practice, although a system (survey) test (SPEI) concept is often used, that is, a fine test is added after a basic test, the basic test is mainly a quasi-rope problem test and a hidden information test is used as an auxiliary test, in a specific case, the basic test of the quality evaluation may be mainly a hidden information test.

Since the target is not specific, the first search (searching) for the hidden information test is to find out the possible target problem of interest of the tested person, and once the target problem is determined, the hidden information test according to the known result can be advanced until the dominance is calculated. The following three situations can occur after testing:

no outstanding item

Where q is 0, meaning that all n +1 terms have no difference in response intensity, i.e., the intensity variation is less than a certain decision threshold, e.g., within 15% of the change in the galvanic skin response intensity (see scoring criteria above), then a score of 0 may be assigned to any of the selected n +1 terms, if desired.

Note that, in this case, q may be regarded as n +1, but the score is not changed.

② has a protrusion

When q is 1, the process can be converted into (1: n) mode.

③ has a plurality of items outstanding

Where q < n +1, it can be decomposed into q groups (1: n) and then processed separately.

In summary, when the q: n situation occurs, in addition to q ═ 0, other tests may be tried in the form of specified items, that is, the attention problem is set and tested again in the form of a known result target problem (q ═ 1), and then the awarding evaluation may be performed.

Example (c): q is 2, then two problems of the reaction occur, and the mapping is performed in the form of (1: n) points.

Generally, the situation of q > 2 in each large quasi-rope problem test rarely occurs, and if the situation occurs, the setting of the subject is examined first to see whether two (or more) types of subjects appear in a group of large quasi-rope problem tests, and the type attribute is based on the standard of the tested person.

In short, when the q: n situation occurs, the situation must be analyzed specifically and processed as appropriate, and the overall processing of the topic content must be considered. The number of geometric averages in the q: n assigning points is determined as follows: firstly, when a situation A appears after the test, taking n +1 to carry out geometric averaging; and secondly, taking n for geometric averaging when situations B and C appear after the test. Any one of the tests or reaction projections can then be assigned a score by comparison with the geometric mean.

Advantage calculation of q: n scoring points

①q＝0

Indicating that all n terms of the intensity of the response are not different, i.e., the intensity variation is less than a certain decision threshold, e.g., the variation of the galvanic skin response intensity is within 15% (see scoring criteria above), and then a score of 0 may be assigned to any one of the selected n terms, if necessary.

According to formula (II) above, when λ ═ 0, P_(-∣T)52.14%, substitution formula (X) is:

L(1:n)-(min)＝0.5214(n+1)/2＝0.2607(n+1)

that is, all terms can now gain some degree of "honest" advantage.

②q＝1

Refer to (1: n) mode processing.

③q＝m

Where m < n +1, it can be decomposed into m groups (1: n) and then processed in (1: n) mode, respectively.

2) Likelihood ratio calculation

The calculation formula of the advantage change (likelihood ratio) of each assigned point measured by the multichannel analyzer is summarized in the following table 4.

TABLE 4

According to the above formula, the atlas probability value obtained by scoring and probability conversion can be converted into a corresponding likelihood ratio, so as to prepare for generation of the joint probability.

Negative results of the covert information test: traditionally, it was thought that covert information tests were used only to confirm criminal use, which resulted in the ignorance and indifference of negative results to the covert information test target problem. The large quasi-rope problem test (BCQT) is proposed to solve the problem of blinding of negative results in covert information tests.

As already mentioned above, when the target problem response strength position of the known result is located at r, and when r is n/2, the target problem response of the covert information test has not had proof power of "cheating", that is, the target problem response has had some proof power of "honest" since then.

Since covert information testing of unknown results can often be converted to known results for processing, the target problem here also includes concerns about the unknown results, which are no longer intentionally distinguished.

Generally, the number of common covert information test problems is 5-7, i.e., n is 4-6, so when the target problem resides in the third reaction intensityThis is essentially the location of the inflection point, and the assigned value is therefore often set to 0, thereby beginning to produce the effectiveness of a negative response.

According to the psychological information "two distributions" result, when λ is 0, P_(-∣T)52.14%, substitution formula (X) is:

L(1:n)-(min)＝0.5214(n+1)/2＝0.2607(n+1)

i.e. the advantage gain in this case is only related to the number of compassing problems.

Theoretically, the more accompanying questions, the higher the likelihood ratio, but the excessive accompanying questions are not only suspected of intentionally confusing the audiovisual work, but also affect the normal generation of the test data, so that the practical operation is controlled to be 4-6, and if necessary, the accompanying questions can be divided into other groups for testing.

It is noted that when the q: n point of awarding is 0, 0.2607(n +1) honest dominance ratio can be obtained for all selected items, so that the unique advantage of the search hidden information test in the quality evaluation is proved, namely, the simultaneous evaluation of a plurality of evaluation items can be realized, and certainly, the basic standard is required for such evaluation items, and the theme setting becomes the key.

First positive result: for a set of covert information tests with n companionic problems and one target problem, when the target problem reaction strength is at r position (when r is 1, the target reaction is at 2 nd position), 1 ≦ r < n, then:

L+＝[(n+1)–r)]P_(+|D)

L–＝(r+1)P_(-∣T)

when the result is negative, the reciprocal of formula (XI) is shown as follows:

L(1:n)－＝L－/L+＝(r+1)P(-|T)/[(n+1)–r)]P(+|D)

since r is the order of the reaction intensity of the target problem, it is a positive (specific) reaction only before n/2, i.e., less than n/2, depending on the purpose of the hidden information test.

The advantage calculation of the hidden information test well explains the difference between the effect of the hidden information test and the effect of the collimation rope problem test, and also lays a foundation for the hidden information test to play a more full role in direction indication and quality evaluation.

The multi-channel instrument test evaluation is composed of a series of assigning points, each assigning point generates own advantage change after testing, and the combination of the advantage changes is the combined advantage.

According to the probability principle and the Bayesian classifier principle, if a certain multi-channel instrument test evaluation comprises k (1:1) assigning points and m (1: n) assigning points, the combined advantages of the test evaluation are as follows:

L_association＝L₀×L_(1∶1)1×L_(1∶1)2×…×L_(1∶1)k×L_(1∶n)1×L_(1∶n)2×…×L_(1∶1)m

In the formula L₀＝1

Obviously, some points of differentiation are forward and some are reverse, and they should be combined separately. Thus, in a multichannel test, the weighted score λ < 0 for the scoring points ("positive" response, scored +) is combined with the weighted score λ ≧ 0 for the scoring points ("negative" response, scored-) respectively, i.e.:

when there are (k +) division points for "(1: 1) division points and (m +) division points for" (1: n) division points, ("1: 1"),

L_{+ Association}＝L₀×L_+(1∶1)1×L_+(1∶1)2×…×L_+(1∶1)k+×L_+(1∶n)1×L_+(1∶n)2×…×L_+(1∶1)m+

When there are (k-) division points of (1:1) and (m-) division points of (1: n) division points with lambda being more than or equal to 0,

L_{-federation of}＝L₀×L_-(1：1)1×L-_(1：1)2×…×L_-(1：1)k-×L_-(1：n)1×L_-(1：n)2×…×L_-(1：1)m-

Wherein (k +) + (k-) ═ k, (m +) + (m-) ═ m, L₀＝1

Specifically, the implementation of the posterior probability comprises the following steps:

calculate proof weight in multichannel tester WoE:

WoE＝L_{+ Association}/L_{-federation of}；

3.1 evidence weight (WoE)

3.1.1 means

Evidence weight means an integration (integration) of information from different sources, according to which the information objective can be achieved. This integration is also a manifestation of bayesian theory.

The value and meaning of the method are that the information quantity of a single piece of evidence cannot meet the information judgment requirement, and each piece of information brings about not only consistency but also opposite evidence.

The Weight (Weight) allows you to determine how useful each evidence is based on various evidence influencing factors, such as data quality, consistency of results, severity of effects, and relevance of information.

Generally, the more information, the more the evidence weighs, but also has a large relationship to your information structure and organization method.

Evidence weights are widely used in classification identification and can be obtained by Basic Odds Ratio (BOR):

BOR＝(Distribution of Good Credit Outcomes)/(Distribution of Bad Credit Outcomes)

or simply:

BOR＝Distr Goods/Distr Bads

when BOR is 1, the positive (supportive) and negative (bad) evidences are in equilibrium, so 1 becomes a judgment threshold. Since it is clear that when BOR > 1, support is stronger than objection, and when BOR < 1, objection is stronger than support.

In general, the BOR is naturally logarithmized to WoE, which is the following formula, but the nature is not changed, but the threshold value is changed from 1 to 0, and more than 1 and less than 1 are positive and negative numbers, respectively.

But to be able to directly correspond to the Bayes Factor (BF), we directly consider BOR here as WoE, i.e. WoE ═ BOR in the qualitative assessment.

Review of

WhereinReferred to as a bayesian factor.

The Bayesian factor quantifies here the certainty that evidence B supports A versus "non-A" (-A), in other words, it quantifies that evidence B supports A with a probability that it is a multiple of the probability that non-A is supported, i.e., BOR. For convenience of use, jeffeys (Jeffreys) originally labeled bayesian factors of different sizes with "significant", "edge significant" and "insignificant" in similar hypothesis testing: evidence that is generally greater than 3 or less than 1/3 is considered substantial (substential evidence); between 1/3 and 3, the weak or evidence to be verified (week or identity evidence). This threshold is not used by the qualification evaluation, but is also of reference value.

In the multi-channel instrument test, the denominator in the above formula is L-association, and the numerator is L + association; the multi-channel tester WoE thus has the following formula:

WoE-L + combination/L-combination

In the crime investigation test, a dominant (L) calculation method is used to determine "guilt" or "innocent" of a person under test according to bayes' theorem.

From the perspective of providing evidence of "guilt," the evidence weight (WoE) at this time may be defined as:

WoE-L guilty coalition/L innocent coalition when:

WoE is less than or equal to 1, the tested person has a test result of 'pass';

when the test result is more than 1 and less than WoE and less than or equal to 5, the test result of the tested person is inconclusive;

when the test result is more than 5 and less than WoE and less than or equal to 9, the tested person is not passed;

WoE > 9, the result of the human test can be used as a basis for "conviction".

Evaluation index for quality (CAI):

the evaluation of merit index (CAI) is intended to more accurately explain the new concept of the meaning and value of evidence weights in the evaluation of merit, and is numerically the inverse of evidence weights.

The quality evaluation directly regards the Bayesian factor as an evidence weight (WoE) or a quality evaluation index (CAI) according to evaluation requirements, so that the probability value conversion between the Bayesian factor and the habit of people is more convenient and faster besides the consideration of understanding, namely, when WoE or CAI needs to be converted into the probability value, the method can be realized by directly adopting the following formula:

therefore, in the Bayesian qualitative assessment,

from the above two equations, it can also be seen that WoE is the probability ratio for event B at conditions (A) and (-A). Note that conventionally, for example, in a crime investigation, one usually considers condition a as a factor contributing to the occurrence of event B, and condition "not a" (-a) as a factor not contributing to the occurrence of event B, so that the more condition a certain evaluation object possesses, the more "crime" (baddie); the more the condition "non-A" (-A) is, the more "innocent". The larger the value of WoE resulting in a determination, the larger the likelihood of a "criminal".

The original intention proposed by the qualitative Assessment is "preferred talent", CAI (qualitative Assessment Index, title abbreviation of creative Assessment Index), which is a core content sought to be obtained by the qualitative Assessment, and can be said to originate from WoE (Evidence Weight, title abbreviation of experience). Since WoE is commonly used to prove the "crime" (miscarriage) degree of a criminal suspect in a criminal investigation, taking its reciprocal can be an indicator to measure the "good knowledge" (happiness) degree of the assessment object. In the category of quality assessment, besides the crime investigation requiring WoE, other cases can be processed in the CAI manner, so that the WoE reciprocal is processed to CAI.

Table 5 is an illustration of the generated qualitative assessment survey.

TABLE 5

The table operates notes:

(1) scoring according to 7 points, wherein the positive point is passed and the negative point is not passed;

(2) inputting the score of each physiological index of each test of each topic into a table; only three tests are selected;

(3) filling the subject contents according to the actually measured subjects, and if the number of the CQT subjects is not enough, copying the CQT format in the line 11, and performing CIT in the same way;

(4) the weights of respiration, blood pressure and skin electricity can be automatically adjusted at the subtotal part of each question;

(5) the weight of each test pass can be adjusted by self at the summation point (lambda) of each topic; when lambda is more than or equal to 0, data should be available only in P-pass/innocent columns, and when lambda is less than 0, data should be available only in P-fail/guilty columns;

(6) the CIT test is limited to the verification test, and the number of the companioning problems is filled in the remark column of the corresponding CIT subject.

In the embodiment of the present invention, the analysis process of the test pattern is explained below with reference to fig. 9.

1. Setting physiological index weight, testing pass weight, false negative rate and false positive rate;

1) setting of physiological index weight, default value: breath is 0.2, electrocardiogram is 0.1, and skin current is 0.7. This can be adjusted empirically, with the sum of the weights of the three indices being 1, with the pico-cell weight recommendation being 0.5-0.7.

2) Setting the weight of the test pass, wherein the default value is as follows: the first pass was 0.2, the second pass was 0.4, and the third pass was 0.4. This can be adjusted empirically, with the sum of the weights of the three indices being 1, suggesting that the first pass test weight does not exceed 0.3.

3) If the set weight is to be applied to all titles, the weight setting operation should be performed before all titles are input with contents and scores. If no weight modification is performed, the system adopts weight default value calculation.

4) If only the weight setting of a topic needs to be changed, the change of the weight should be performed after the content of the topic and the score are input. And again before the next topic is entered, which would otherwise affect all subsequent topics, suggesting separate processing for each topic.

5) The false negative rate and the false positive rate are set for each relevant topic empirically, and the maximum value is 0.5.

Setting conditions: when the result of the tested person on a relevant question is 'pass' (namely the weighted sum of the question scores is more than 0), the false negative rate adjustment is meaningful; the false positive rate adjustment is meaningful when the result of the tested person on a relevant question is "fail" (i.e. the weighted sum of the question scores is less than 0).

2. Selecting a topic mode, and inputting topic contents and topic scores;

1) selecting a topic mode, and defaulting to a CQT mode. A 7-part method was chosen for scoring. The CIT topics need to input n values (number of accompanying topics).

2) After inputting the subject content, selecting the score values of the three-time test, inputting the score values according to the order of the subjects, and storing the score values. Until all the topics are input.

3) If the item input is wrong, the item can be searched and modified before the program is closed, and then the item is stored again.

The method comprises the following specific steps:

clicking on "find last question" will display the last question that has been saved. Continuing to search for the previous question, jumping to and displaying the last but one question, and so on. After each item of information of the current topic is modified, clicking 'storage \ modification topic' will modify all the information of the topic and store the modified information in the memory.

3. And setting output file names and storage path paths of all titles.

1) Setting file name and output path

The output file type of the calculation result is a text file. Txt, the file name can be modified according to actual conditions.

2) Clicking 'calculation' can display the advantage value and the test result

After setting the contents, scores, weights, false negative rates (false positive rates) and file names of all titles is completed, the storage path of the file is selected, i.e., "calculate" is clicked, and a value of WoE is displayed.

3) Finally, a ". txt" file is generated.

The Bayesian theory-based system testing and map analyzing method provided by the embodiment of the invention has the following beneficial effects:

(1) testing method characteristics and functions

The purpose of finding out the accurate psychological information reaction point of the tested person (or suspect) by a test mode of common situation common theory based on information exploration as guidance, psychological information as a core and information coupling can be realized by a well-organized system (survey) test, and an evidence method can be helped to be established. The evidence method is not only beneficial to improving the knowledge of a judicial staff on the multi-channel instrument test, but also can improve the evidence idea of the common people, and can play a larger role on the basis of the original investigation means.

Mainly embodied in the following aspects:

verification of

The verifiability of the fine test is intended to verify that the person under test determined to be related to the case (event) under evaluation by the basic test is such that a level of confidence is achieved. Of course, sometimes the verification can also be embodied in determining whether the tested person has a correlation with the evaluated case (event).

Exploratory property

In the investigation test function, the exploratory test is often used as a detection means, which is helpful for determining the detection direction, finding the evidence, finding the falling of dirt, and the like. In a screening test or an inspection test, exploratory properties often help to further ascertain details of the person under test related to the case (event) under evaluation.

Expansibility

In the criminal investigation test, expansibility means that whether other problems exist besides the fact that a tested person is related to an investigated case or not is solved by using a fine test, the test is set for expanding the fighting results, and a plurality of non-beginners of the same type or a type of criminal case are often tested by using an expansibility test theme in actual combat so as to achieve psychological information required for deeply digging a crime.

Non-uniformity of search

The verification, searchability and extensibility are not achieved for the fine testing of each human being tested. In practice, different types and different quantities of fine test subjects are often set according to the specific situation of a case, and the types and the quantities of the test subjects are flexibly determined and used according to the progress situation of the test.

Systematicness

The basic test and the fine test are not simple and isolated, have inherent connectivity, and need to be carefully set according to the characteristics of people and correspond to each other. The decision as to whether to perform the fine test is made only after the basic test results are obtained, and if it is proved that the person under test "passes" the basic test, the fine test is not performed in general. While when the person under test "fails" the basic test, the fine test must be performed and is usually a spread and drill-in of problems associated with the basic test.

Robustness

Practice shows that the systematic (investigation) test has very good robustness (robustness), i.e. although it is designed for criminal investigation test, its utility is also very suitable for screening test and inspection test, and becomes the solid foundation of multi-channel tester for quality evaluation.

(2) Innovativeness of atlas analysis method

2.1 idea of two distributions

The two distributions provide a base point for scientific judgment of the multichannel instrument test, and the base point becomes a basic reference of the whole system, and the functions of the two distributions are represented as follows: the test method has the advantages that the overall accuracy of the test can be accurately evaluated; an ROC curve is drawn through a signal detection theory, and the overall accuracy of the test can be measured; secondly, providing a basic basis for atlas evaluation and analysis; data support is provided for setting a scoring standard, so that scoring is more rigorous and accurate; the third indicates the direction for the technical improvement; the technical limitation and the advantages are accurately found, and the technical level is improved and improved in a targeted manner.

Depending on the statistics chosen, there should also be multiple types of "two distributions".

2.2 application of Bayesian theorem to dominant expression

In multi-channel testing, bayesian decision making is implemented by the expression of the superiority of bayesian theorem (likelihood ratio L). The advantage expression of Bayes ' theorem not only avoids the influence of the prior probability thoroughly from the mathematical form, but also has the maximum contribution that the joint advantage can be obtained through the joint probability, so as to obtain the evidence weight, and the promotion and the conversion from ' uncertain ' to ' determined ' are completely realized through the integration of all information points. Relative to the accuracy of the past, the odds ratio (or L) is used to judge or is easier to understand and master by using the change of the prior probability and the posterior probability.

2.3 lifting of the Large guide rope

The large guideline is a name for regarding all accompanying lining (background) problems in the hidden information test as the guideline problem, and after the concept is introduced into the multi-channel instrument test, the barrier of data analysis of the guideline problem test and the hidden information test in the traditional sense is broken, and simultaneously, the barrier is cleared for the application of Bayesian theorem in the data analysis of the multi-channel instrument test.

In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention, and that variations, modifications, substitutions and alterations can be made in the above embodiments by those of ordinary skill in the art without departing from the principle and spirit of the present invention. The scope of the invention is defined by the appended claims and equivalents thereof.

Claims (10)

1. A system testing and map analyzing method based on Bayesian theory is characterized in that a multichannel instrument is adopted for testing, and the method comprises the following steps:

step S1, according to the present evaluation event, the evaluation object is basically tested, if the basic test is passed, the test conclusion is generated, otherwise, the step S2 is executed;

step S2, performing fine test on the evaluation object, and generating a test conclusion no matter whether the fine test is passed or not;

wherein, the basic test and the fine test both adopt one or more of a quasi-rope problem test method and a hidden information test method.

2. The bayesian-theory-based system testing and mapping method according to claim 1, wherein a combination of two sets of multi-objective quasi-rope problem unit tests and one set of hidden information unit tests is set in the basic test of the step S1.

3. The Bayesian theory-based system testing and pattern analysis method of claim 2, wherein the question structure of the quasi-string problem testing method requires that each group of test questions must include three types of test questions, namely, related questions, quasi-string questions and unrelated questions.

4. The bayesian-theory-based system testing and mapping method according to claim 2, wherein the hidden-information testing method comprises: the key question and the accompanying question are used for detecting whether the psychological pressure responses of the evaluation object to the key question and the accompanying question are different or not, and further deducing whether the psychological information of the evaluation object contains the question elements to be investigated or not.

5. The bayesian-theory-based system testing and pattern analysis method of claim 1, wherein in the multi-channel tester testing, the pattern probability in the test conclusion is converted into the posterior probability by bayesian theorem.

6. The Bayesian theory-based system testing and mapping method according to claim 5, wherein in the quasi-rope problem testing, physiological index data are collected and analyzed in a mapping manner.

7. The Bayesian theory-based system testing and mapping method of claim 6, wherein the physiological index data comprises:

(1) respiration takes the breath line length as a measure of Ir and Ic;

(2) skin electrification adopts peak height or peak area;

(3) the blood pressure is raised by the base line, and lowered or increased by the base line.

8. The Bayesian theory-based system testing and mapping method according to claim 5, wherein the posterior probability realization comprises the following steps:

calculate proof weight in multichannel tester WoE:

WoE＝L_{+ Association}/L_{-federation of}；

When WoE or CAI is converted into probability value, the method can be realized by directly adopting the following formula:

therefore, in the Bayesian qualitative assessment,

where WoE is the probability ratio for event B at conditions (A) and (-A).

9. The bayesian-theory-based system testing and profile analyzing method according to claim 8, wherein in the posterior probability implementation, two normal distributions are set according to scores in a regional comparison test by a bexter scoring technique, and an ROC curve is drawn based on a signal detection theory.

10. The bayesian-theory-based system testing and pattern analysis method according to claim 5, wherein analyzing the test pattern comprises:

setting physiological index weight, testing pass weight, false negative rate and false positive rate;

selecting a topic mode, and inputting topic contents and topic scores;

and setting output file names and storage paths of all titles.