Special fraud prevention system, special fraud prevention program, and special fraud prevention method

Abstract

To provide a special fraud prevention system, a special fraud prevention program, and a special fraud prevention method which appropriately call attention to a person who seems to encounter a special fraud suffering, and can prevent special fraud.SOLUTION: A special fraud prevention system 1 is a system for activating an alarm to a target person having the possibility to encounter a special fraud suffering, and sets an alarm level indicating the possibility of the special fraud suffering, from a common factor calculated by factor analysis based on an evaluation value of previously prepared questionnaire information on a feature of a sufferer, and a person feature amount based on image analysis of the target person. Thereby, the special fraud prevention system can enhance a detection rate of the target person encountering a special fraud suffering, and can prevent a special fraud suffering.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a special fraud prevention system, a special fraud prevention program, and a special fraud prevention method, and more particularly to a special fraud prevention system, a special fraud prevention program, and a special fraud prevention method that can appropriately warn people who are likely to become victims of special fraud and prevent special frauds from occurring. [Background technology]

[0002] In recent years, there has been an increase in special frauds, including bank transfer fraud, refund fraud, and fictitious billing fraud, and the methods used to commit these crimes have also become more diverse. For example, in bank transfer fraud, criminal groups use some pretext to try to extract money from elderly people. The methods vary, including transferring the money through an automated teller machine (ATM) and having the money collected by a receiver. In this situation, police and banks have been stationing staff around ATMs installed within banks to monitor people who are likely to be targets of bank transfer fraud (such as the elderly) and to warn them by calling out to them.

[0003] As a prior art for preventing bank transfer fraud, for example, Patent Document 1 discloses a technology that has a photographing device that photographs a customer, analyzes the customer's image, determines whether the customer is on a call using a mobile phone, and if the customer is making a transfer transaction and is on a call, displays a screen urging the customer to stop the call or stop the transaction and go to a counter, thereby preventing damage caused by bank transfer fraud from occurring.

[0004] Furthermore, Patent Document 2 discloses a technology that has a microphone that picks up sounds near the ATM, analyzes the audio picked up by the microphone, and determines that the audio is one-way audio if the audio intervals of the picked up audio are separated by more than a predetermined time and are repeated a predetermined number of times, and in this case warns the ATM user that there is a risk of transfer fraud. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Patent No. 6959704 [Patent Document 2] Japanese Patent Application Laid-Open No. 2010-079741 Summary of the Invention [Problem to be solved by the invention]

[0006] However, the technologies disclosed in the aforementioned Patent Documents 1 and 2 are based on the premise that the victim is making a call using a mobile phone, and therefore, depending on the method used in the bank transfer fraud, it may be difficult to accurately identify the perpetrator of the crime.

[0007] For example, there are cases where a person who is suspected to be a victim of a bank transfer fraud is talking on a mobile phone away from an ATM, but stops talking on the mobile phone when he or she gets close to the ATM, or where the person receives instructions to make a transfer before entering a branch such as a bank, and then enters the branch and operates the ATM without speaking to anyone. Depending on the modus operandi of such bank transfer fraud, it may not be possible to prevent the occurrence of bank transfer fraud even if the technology in Patent Document 1 or Patent Document 2 is used.

[0008] The present invention has been devised in consideration of the above points, and aims to provide a special fraud prevention system, a special fraud prevention program, and a special fraud prevention method that can appropriately warn people who are likely to become victims of special fraud and prevent special fraud from occurring. [Means for solving the problem]

[0009] In order to achieve the above-mentioned object, the special fraud prevention system of the present invention comprises a database in which questionnaire information corresponding to a plurality of questions regarding the characteristics of victims of special fraud, a plurality of answer options to the questions, and evaluation values ​​for the options is stored for each respondent; a person feature calculation unit that calculates person feature values ​​including the target person's movements and appearance through image analysis; a common factor determination unit that determines a first common factor, which is a common factor with similar tendencies, through factor analysis based on the evaluation values ​​of the questionnaire information, and determines a second common factor, which is a common factor from the first common factors that best matches the characteristics of the target person, based on the person feature values; and an alarm level setting unit that sets an alarm level indicating the possibility of being a victim of special fraud based on the second common factor.

[0010] Here, by providing a database in which multiple questions regarding the characteristics of victims of specialized frauds, multiple answer options for those questions, and questionnaire information corresponding to the evaluation values ​​for those options are stored for each respondent, it is possible to accumulate questionnaire information from experts and field personnel (hereinafter referred to as "experts, etc.") who are bank employees or police officers who are familiar with the circumstances surrounding victims of specialized frauds, and to investigate and analyze the know-how about specialized frauds that these experts, etc. possess, and determine the possibility of being a victim of specialized fraud in accordance with the judgment criteria of the experts, etc.

[0011] Furthermore, by providing a person feature calculation unit that calculates person feature amounts, including the movements and appearance of a target person through image analysis, it is possible to, for example, capture an image of an ATM operator using a surveillance camera installed in a bank, and calculate the person feature amounts of the operator through image analysis. Then, based on the calculated person feature amounts, it is possible to determine whether the target person is likely to become a victim of special fraud.

[0012] In addition, by providing a common factor determination unit that determines the first common factor, which is a common factor with similar trends, through factor analysis based on the evaluation values ​​of the questionnaire information, it is possible to identify common factors of special fraud victims that are in line with the judgment criteria of experts, etc., by performing factor analysis on the questionnaire information.

[0013] In addition, the common factor determination unit determines, based on the person features, a second common factor, which is a common factor from among the first common factors that best matches the features of the target person, and can thereby identify, from among the first common factors, a common factor that best matches the person features.

[0014] In addition, by providing an alarm level setting unit that sets an alarm level indicating the possibility of being a victim of special fraud based on the second common factor, the alarm level can be set taking into consideration the characteristics of the target person, such as whether or not they make calls, their behavior, gender, age group, etc., thereby making it possible to prevent special fraud from occurring with a high degree of accuracy.

[0015] Furthermore, when the common factor determination unit determines the common factor with the largest determination value calculated based on the factor loading based on the evaluation value of the first common factor and the person features as the second common factor, it can determine the most influential common factor among the common factors extracted from the first common factor based on the person features as the second common factor, thereby making it possible to set an alarm level according to the characteristics of the target person with high accuracy.

[0016] Furthermore, when the alarm level setting unit changes the alarm level setting based on the ratio of the judgment value to the maximum possible factor loading of the second common factor, the alarm level can be set based on the matching rate between the target person and the common factor, thereby enabling the alarm level to be set with high precision according to the characteristics of the target person.

[0017] Furthermore, when the common factor determination unit determines the first common factor by applying a predetermined rotation to the coordinate axis for the factor loading based on the evaluation value of the questionnaire information, it makes the relationship between the extracted first common factor and the variables apparent, making it possible to easily identify factors that may cause special fraud damage.

[0018] Furthermore, when the common factor determination unit determines the first common factor based on the factor loadings normalized by the square root of the sum of squares of the factor loadings based on the evaluation values ​​of the questionnaire information, the contribution rate of each common factor is leveled out, thereby preventing a specific factor from standing out in the determination criteria and improving the accuracy of the factor analysis.

[0019] In order to achieve the above-mentioned object, the special fraud prevention program of the present invention is designed to have a computer execute the following steps: determining a first common factor, which is a common factor with similar tendencies, through factor analysis based on the evaluation values ​​of survey information stored for each respondent, which includes multiple questions regarding the characteristics of victims of special fraud, multiple answer options to the questions, and survey information corresponding to evaluation values ​​for the options; calculating personal features including the movements and appearance of the target person through image analysis; determining a second common factor, which is a common factor among the first common factors that best matches the characteristics of the target person, based on the personal features; and setting an alarm level indicating the possibility of being a victim of special fraud based on the second common factor.

[0020] In addition, the method for preventing special fraud of the present invention includes a step of determining a first common factor, which is a common factor with similar tendencies, by factor analysis using evaluation values ​​of questionnaire information stored for each respondent, which includes a plurality of questions regarding the characteristics of victims of special fraud, a plurality of answer options to the questions, and questionnaire information corresponding to evaluation values ​​for the options; a step of calculating personal features including the movements and appearance of the target person by image analysis; a step of determining a second common factor, which is a common factor from the first common factors that best matches the characteristics of the target person, based on the personal features; and a step of setting an alarm level indicating the possibility of victimization of special fraud based on the second common factor.

[0021] Here, a step (process) is provided in which a first common factor, which is a common factor with similar tendencies, is determined by factor analysis based on the evaluation values ​​of the survey information stored for each respondent, which corresponds to multiple questions regarding the characteristics of victims of special frauds, multiple answer options for the questions, and evaluation values ​​for the options.By factor analyzing the survey information from experts, etc., it is possible to identify common factors of victims of special frauds that meet the judgment criteria of experts, etc.

[0022] Furthermore, by providing a step (process) of calculating the target person's behavior and personal characteristics, including their appearance, through image analysis, it is possible to, for example, capture an image of an ATM operator using a surveillance camera installed in a bank, and calculate the operator's personal characteristics through image analysis. Based on the calculated personal characteristics, it is then possible to determine whether the target person is likely to become a victim of special fraud.

[0023] Furthermore, by providing a step (process) of determining, based on the person features, the second common factor, which is the common factor from the first common factors that best matches the characteristics of the target person, it is possible to identify the common factor from the first common factors that best matches the person features.

[0024] In addition, by including a step (process) of setting an alarm level indicating the possibility of being a victim of special fraud based on the two common factors, the alarm level can be set taking into account the characteristics of the target person, such as whether or not they have made a call, their behavior, gender, age group, etc., thereby making it possible to prevent special fraud from occurring with a high degree of accuracy. [Effects of the Invention]

[0025] The special fraud prevention system, special fraud prevention program, and special fraud prevention method according to the present invention are capable of appropriately warning people who are likely to become victims of special fraud and preventing special fraud before it occurs. [Brief explanation of the drawings]

[0026] [Figure 1] 1 is a diagram showing a state in which a special fraud prevention system according to an embodiment of the present invention is connected to each device via an Internet line. [Figure 2] FIG. 2 is a block diagram showing the internal configuration of the special fraud prevention system. [Figure 3] FIG. 10 is a diagram showing an example of the contents of a questionnaire for experts and the like. [Figure 4] FIG. 10 is a diagram showing an example of a data format of questionnaire information stored in a database. [Figure 5] 10A and 10B are diagrams illustrating calculation results by a person feature amount calculation unit. [Figure 6] A diagram showing the processing flow of the special fraud prevention program. [Figure 7] FIG. 10 is a diagram showing a matrix AS obtained by calculating the initial values ​​of factor loadings. [Figure 8] FIG. 10 is a diagram showing two-dimensional coordinate axes when a rotation R is applied to a matrix AS. [Figure 9] FIG. 10 is a diagram showing a matrix A after a rotation R is applied to the matrix AS. [Figure 10] FIG. 10 is a diagram showing a matrix A0 obtained by normalizing a matrix A. DETAILED DESCRIPTION OF THE INVENTION

[0027] Hereinafter, embodiments of the present invention relating to a special fraud prevention system, a special fraud prevention program, and a special fraud prevention method will be described with reference to the drawings to help understand the present invention.

[0028] First, an overview of a special fraud prevention system 1 according to an embodiment of the present invention will be described with reference to Figure 1. The special fraud prevention system 1 according to an embodiment of the present invention is a computer for running a special fraud prevention program, which will be described later. The special fraud system 1 can be a server installed in a financial institution such as a bank, or a rental server (cloud server) on which the special fraud prevention program is installed.

[0029] The special fraud prevention system 1 is composed of an ATM terminal 2 installed in a financial institution and an imaging device 3 that monitors the ATM corner where the ATM terminal 2 is installed and captures and records real-time moving images, which are connected to each other via an internet line 4 so that they can communicate with each other.

[0030] Here, the special fraud prevention system 1, the ATM terminal 2, and the imaging device 3 do not necessarily have to be connected to each other via the Internet line 4, but may be connected by any communication means.

[0031] Each component of the special fraud prevention system 1 will be explained in detail below with reference to Figure 2. The special fraud prevention system 1 is mainly composed of a database (DB) 10 in which various data is stored, a person feature amount calculation unit 20 that calculates person feature amounts by image analysis of moving images captured by the imaging device 3, a common factor determination unit 30 that determines predetermined common factors by factor analysis, and an alarm level setting unit 40 that sets an alarm level that indicates the possibility of special fraud victimization. The detailed configuration of the special fraud prevention system 1, including these main components, will be explained below.

[0032] [Database] The database 10 is a non-volatile memory that stores the OS, application software such as special fraud programs, setting data, etc. Based on past cases of special fraud, the database 10 contains questions about the characteristics of ATM operators (hereinafter referred to as "operators") who are likely to become victims of special fraud, and stores the results of a questionnaire survey of experts and others on the importance of each characteristic for each respondent.

[0033] An example of the contents of a questionnaire distributed to experts and the like is shown in Figure 3. Five questions are prepared: "1. Elderly," "2. Operating while talking on the phone or looking at notes," "3. Female," "4. Behavior is unnatural," and "5. Negative facial expression." Experts and the like are asked to respond to these questions on a five-point scale from 1 to 5, and the collected survey results for a number of people (M) (in the embodiment of the present invention, M=1000) are stored in database 10 for each respondent in a predetermined data format. More specifically, as shown in Figure 4, respondents are entered in the row direction of a matrix, and each respondent's evaluation value for each question is entered in the column direction on a five-point scale.

[0034] Here, the questions in the survey information are not necessarily limited to the five mentioned above, and more questions may be added or the number of questions may be limited to fewer than five. Also, the questions may be appropriately revised in consideration of the type of special fraud victim, and the survey information may be updated each time. Furthermore, the number of respondents may also be changed as appropriate.

[0035] [Person feature calculation section] The personal feature amount calculation unit 20 analyzes the facial expressions and actions of the operator operating the ATM terminal 2 from the video images captured by the imaging device 3, and quantifies the operator's features. It then has the function of determining which of the questions in the questionnaire information the operator matches, thereby setting an alarm notification level based on the operator's personal features.

[0036] First, the video captured by the imaging device 3 is temporarily stored in a database 10 via an internet line 4. Then, based on frame images separated from the video stored in the database 10, a person feature amount calculation unit 20 analyzes at least one or more biometric characteristics of the operator, such as facial expression, gender, behavior, and belongings. A trained machine learning model is used for these biometric analyses.

[0037] For example, in facial expression analysis, specific frame images extracted from video are used as input values, the operator's facial area is identified from the frame images, and the identified facial expressions are classified into multiple categories according to a machine learning model. Then, based on the classification results, an analysis is performed based on training data to determine whether there is a positive or negative change in facial expression between consecutive frame images, as well as the extent of the change in facial expression, and the results are output.

[0038] In addition, in analyzing gender, just like analyzing facial expressions, the area of ​​the operator's face is identified from the frame image, and features such as the eyes, nose, mouth, and contours are compared with the training data to identify the operator's gender and output the results.

[0039] Additionally, behavior analysis involves analyzing the operator's behavior through posture estimation. Specifically, the positions of the operator's joints and feature points such as the eyes and nose are estimated from the frame images input into the machine learning model. Then, by acquiring coordinate point clouds of the shoulders, elbows, wrists, waist, knees, ankles, nose, and eyes in time series, the system can detect behaviors such as looking around restlessly or suddenly crouching down, and output the presence or absence of unnatural behavior.

[0040] The biometric analysis method by the person feature amount calculation unit 20 described above is an example, and the biometric analysis does not necessarily have to be performed by the above analysis method, but any known biometric analysis method can be appropriately selected and applied.

[0041] An example of the calculation results calculated by the person feature amount calculation unit 20 for a specific operator is shown in Fig. 5. Through the biometric analysis described above, the person feature amount calculation unit 20 inputs "1" if the question item in the questionnaire information applies, and "0" if the question item does not apply, and in this embodiment of the present invention, it is assumed that the matrix X0 shown in Fig. 5 is obtained. The matrix X0 is temporarily stored in the database 10 and is used in the calculation process for setting the alarm level together with the determination result by the common factor determination unit 30, which will be described later.

[0042] [Common factor determination section] The common factor determination unit 30 has a function of extracting common factors from the questionnaire information by factor analysis, and first determines the number of common factors (first common factors) with similar tendencies based on the questionnaire information. Furthermore, based on the first common factors and the operator's characteristics calculated by the person feature amount calculation unit 20, it determines the common factor (second common factor) that best matches the operator's characteristics. The factor analysis method used by the common factor determination unit 30 will be described in detail later.

[0043] [Alarm level setting section] The alarm level setting unit 40 has the function of setting an alarm level indicating the possibility of being a victim of special fraud, based on the judgment result of the common factor judgment unit 30, taking into consideration the target person's characteristics such as whether or not they have made a call, their behavior, gender, age group, etc.

[0044] The alarm level can be set to multiple stages based on the match rate between the second common factor and the operator determined by the common factor determination unit 30. For example, if the match rate between the second common factor and the operator is less than 40%, no warning is issued, and if the match rate is 40 to 80%, a notice is displayed on the ATM screen indicating that there is a risk of special fraud. Furthermore, if the match rate is 80% or higher, it is determined that there is an extremely high possibility of special fraud, and in addition to the notice on the ATM screen, a notice is sent to the responsible person.

[0045] In this way, by notifying operators of the possibility of becoming victims of special fraud, it is possible to raise awareness among operators and managers about the possibility of becoming victims of special fraud, and to prevent such fraud from occurring in the first place.

[0046] The alarm level threshold set by the alarm level setting unit 40 can be changed arbitrarily. In the example described above, the alarm level was changed in three stages, but the alarm level can be set in four stages, or more, or by providing two threshold stages. Furthermore, the method of notifying the operator or manager of the alarm can also be changed as appropriate.

[0047] The above is the configuration of the special fraud prevention system 1 according to the embodiment of the present invention. Next, the special fraud prevention program executed by the special fraud prevention system 1 will be described with a specific example.

[0048] The special fraud prevention program is executed according to the flow diagram in Figure 6. That is, a first common factor is determined by factor analysis based on the evaluation values ​​of the questionnaire information (S1), then the personal feature calculation unit 40 calculates the personal feature of the operator (S2), and based on the first common factor and the personal feature, the factor among the first common factors that best matches the characteristics of the operator is determined to be a second common factor (S3), and further an alarm level is set based on the second common factor (S4).

[0049] The factor analysis performed in the special fraud prevention program will be explained in detail below using a specific example.

[0050] <Correlation matrix calculation> First, from the matrix X of the survey information shown in Figure 4, the correlation matrix K (K = [k nn ])

[0051] Also, a normal random matrix X of the same size as matrix X r and also generate the correlation matrix K r (K=[k rnn ]) is calculated. nn , and k rnn is an element of the set of real numbers.

[0052] <Eigenvalue calculation> From the correlation matrix K, the eigenvalue vector λ (λ = [λ n ]) where λ is assumed to be in descending order.

[0053] Similarly, the correlation matrix K r from the eigenvalue vector λ r (λ r =[λ rn ]). λ r are also arranged in descending order, just like λ.

[0054] <Determining the number of factors> Next, λ > λ r The number of factors T, i.e., the number of common factors obtained from the survey information, is determined. Here, if the tendency of the survey results by the respondents can be summarized into one answer, or if there is no consistency in the answer trends, T=1. However, in the embodiment of the present invention, T=2 is assumed and used in the following calculations.

[0055] <Calculation of initial values ​​of factor loadings> Once the number of factors is determined, the factor loading matrix A that satisfies the relationship in the following formula (1) is created. S Here, matrix F is a common factor and matrix E is a unique factor. In the embodiment of the present invention, as a result of calculation based on [Equation 1], matrix A shown in FIG. S Let us assume that the matrix A is obtained. S Each numerical value is the initial value of the factor loading for each questionnaire item of the first and second factors.

[0056]

number

[0057] <Determining the first common factor> After calculating the factor loadings for each factor, we determine the first common factor, which is a specific common factor that serves as a criterion for determining special fraud common to the survey respondents.To determine the first common factor, we first plot each factor loading on a two-dimensional coordinate system, with the first factor determined by calculating the initial factor loadings on the horizontal axis and the second factor on the vertical axis.At this time, we perform factor analysis with promax rotation to make it easier to interpret the factor loadings for each factor.

[0058] Promax rotation is an oblique rotation method used in factor analysis, and allows for correlations between factors. Instead of Promax rotation, factor analysis may be performed with Varimax rotation, which is an orthogonal rotation method used in factor analysis.

[0059] Figure 8 shows the matrix A S The two-dimensional coordinate axes are shown when a rotation R is applied to the matrix A. As shown in Figure 8, the above factor analysis allows us to set the first factor as "operating while talking or looking at notes (talking / notes)" and the second factor as "elderly person." Then, the matrix A S The matrix A after applying the rotation R is shown in FIG.

[0060] <Normalization of factor loadings for the first common factor> Once the first and second factors have been determined, matrix A0 is obtained by normalizing matrix A using the square root of the sum of squares so that the contribution rate of each factor is equal. The normalized matrix A0 is shown in Figure 10.

[0061] Note that normalization of matrix A is not necessarily required, but as shown in Figure 10, the normalized matrix A0 prevents the first factor (calls and notes) from standing out as a judgment criterion compared to the second factor (elderly) in comparison to matrix A. This prevents specific common factors from becoming judgment criteria when setting alarm levels, enabling highly accurate detection of the possibility of special fraud victimization.

[0062] <Determining the second common factor> After the first common factor is determined, the second common factor is determined based on the first common factor and the person feature calculated by the person feature calculation unit 20.

[0063] First, based on the matrix A0 and the matrix X0 calculated by the person feature amount calculation unit 20, the degree of match y0 between the first factor (call memo) and the second factor (elderly person) is calculated using equation (2).

[0064]

number

[0065] Then, when the degree of agreement between the first and second factors is calculated based on the formula [2], the degree of agreement for the first factor is 0.29 (= 0.29 + 0.1 - 0.1), and the degree of agreement for the second factor is 1.64 (= 0.70 + 0.47 + 0.47), and the second factor has a larger value than the first factor (the degree of agreement for the second factor at this time is called "y max Therefore, in the embodiment of the present invention, the second factor is determined as the second common factor because it matches the second factor best.

[0066] <Alarm level setting> Second common factor and agreement y max Once this is determined, the maximum possible factor loading (Y max ) is found.

[0067] Y max is the sum of all positive factor loadings of the second factor in matrix A0 shown in Figure 10 (0.7 + 0.12 + 0.47 + 0.47), and Y max =1.76. Then, the alarm level setting unit 40 sets the alarm level ρ according to the formula [3].

[0068]

number

[0069] The alarm level ρ calculated based on the formula 3 in this embodiment of the present invention is 93% (=1.64 / 1.76*100). This alarm level corresponds to a match rate of 80% or more when judged based on the threshold value described above, so in this case, in addition to notifying the operator by a notification on the ATM screen, a notification is also sent to the person in charge.

[0070] By repeating the above calculations, it is possible to set the most effective alarm for the target operator and notify them of the possibility of being a victim of special fraud.It also works effectively against special frauds that do not require conversation, thereby improving the accuracy of detecting special frauds.

[0071] As described above, the special fraud prevention system, special fraud prevention program, and special fraud prevention method of the present invention are capable of appropriately warning people who are likely to become victims of special fraud and preventing special fraud from occurring. [Explanation of symbols]

[0072] 1. Special fraud prevention system 2 ATM terminals 3. Imaging device 4. Internet connection 10 Databases 20 Person feature calculation unit 30 Common factor determination section 40 Alarm level setting section

Claims

1. a database in which questionnaire information is stored for each respondent, in which a plurality of questions regarding characteristics of victims of special fraud, a plurality of answer options for the questions, and evaluation values ​​for the options are associated with each other; a person feature amount calculation unit that calculates person feature amounts including the movement and appearance of a target person through image analysis; a common factor determination unit that determines a first common factor, which is a common factor having a similar tendency, by factor analysis based on the evaluation values ​​of the questionnaire information, and determines a second common factor, which is a common factor that best matches the characteristics of the target person among the first common factors, based on the person feature amount; an alarm level setting unit that sets an alarm level indicating the possibility of being a victim of special fraud based on the second common factor. Special fraud prevention system.

2. The common factor determination unit determines, as the second common factor, a common factor for which a determination value calculated based on a factor loading based on an evaluation value of the first common factor and the person feature amount is maximum. The special fraud prevention system according to claim 1.

3. The alarm level setting unit The setting of the alarm level is changed based on the ratio of the judgment value to the maximum value that the factor loading of the second common factor can take. The special fraud prevention system according to claim 2.

4. The common factor determination unit The first common factor is determined by applying a predetermined rotation to the coordinate axis of the factor loading based on the evaluation value of the questionnaire information.

3. The special fraud prevention system according to claim 1 or 2.

5. The common factor determination unit The square root of the sum of squares of the factor loadings based on the evaluation values ​​of the questionnaire information is calculated, and the first common factor is determined based on the factor loadings normalized by the square root of the sum of squares.

3. The special fraud prevention system according to claim 1 or 2.

6. A step of determining a first common factor, which is a common factor with similar tendencies, by factor analysis based on the evaluation values ​​of questionnaire information stored for each respondent, in which multiple questions regarding the characteristics of victims of special fraud, multiple answer options for the questions, and evaluation values ​​for the options are associated; calculating a person feature amount including the movement and appearance of the target person by image analysis; determining a second common factor, which is a common factor that best matches the characteristics of the target person, from among the first common factors based on the person feature amount; and setting an alarm level indicating the possibility of being a victim of special fraud based on the second common factor. Special fraud prevention program.

7. The step of determining the second common factor includes a step of determining, as the second common factor, a common factor for which a determination value calculated based on a factor loading based on an evaluation value of the first common factor and the person feature amount is maximum. The special fraud prevention program according to claim 6.

8. The step of setting the alarm level includes: and changing the setting of the alarm level based on the ratio of the judgment value to the maximum value that the factor loading of the second common factor can take. The special fraud prevention program according to claim 7.

9. A step of determining a first common factor, which is a common factor with similar tendencies, by factor analysis using the evaluation values ​​of the questionnaire information stored for each respondent, in which multiple questions regarding the characteristics of victims of special fraud, multiple answer options for the questions, and evaluation values ​​for the options are associated; calculating a person feature amount including the movement and appearance of the target person by image analysis; determining a second common factor, which is a common factor that best matches the characteristics of the target person, from among the first common factors based on the person feature amount; and setting an alarm level indicating the possibility of special fraud damage based on the second common factor. Special fraud prevention methods.

Images