An anti-fraud volunteer intelligent matching system, method and device

Abstract

The present invention claims protection for an anti-fraud volunteer intelligent matching system, method and device, which belongs to the field of intelligent volunteer management technology. Intelligent matching includes: a personnel event intelligent matching model, a login module, an application scenario management module, a process engine, an event knowledge manual, an element overview module, a personnel management and organization management module, a regional management module, an event management module and an equipment management module. The server receives the help information sent by the help seeker via the wireless communication module based on NB-IoT; according to the application scenario management module, the category of volunteers required by the help seeker is determined according to the help information; according to the pre-set volunteer discrimination classification rules and the volunteer category, the anti-fraud volunteers are matched in the volunteer information database, and the matched anti-fraud volunteers are obtained according to the personnel management and organization management module; according to the regional management module, the help information is forwarded to the terminal of the matched volunteer for processing. The present invention can improve the accuracy of volunteer matching and improve the management efficiency of personnel.

Description

Technical Field

[0001] The present invention belongs to the field of volunteer management, and specifically belongs to the technical field of intelligent matching of volunteers. Background Art

[0002] With the rapid development of mobile Internet technology and mobile terminals, the original volunteer matching method has the following technical problems:

[0003] 1. Difficulty in managing scene personnel

[0004] Problem to be solved: There is no clear management for participants in scenarios such as anti-fraud and joint patrols. Due to the lack of effective management, many grassroots staff may handle tasks in various scenarios at the same time, resulting in inefficiency.

[0005] 2. The processing flow is not visible

[0006] Problem to be solved: There is a simple large-node process for anti-fraud scenarios (receiving warning - telephone dissuasion - door-to-door dissuasion - feedback), but there is no response process for different situations, and the process details are not handled at each major node. For example, at the telephone dissuasion and door-to-door dissuasion nodes, there is no detailed process details based on different time, identity of the parties, event type, event address and other information. When the door-to-door dissuasion effect is not good, such as the door-to-door dissuasion effect of grid members or volunteers for students is not as good as that of teachers in the school, there is no data accumulation on the dissuasion effect, resulting in a lag in the adjustment and optimization of the processes in the nodes.

[0007] 3. Inappropriate handling personnel

[0008] Problem to be solved: Some door-to-door dissuasion personnel in anti-fraud scenarios are inexperienced. Due to lack of experience and lack of understanding of process details, sometimes it is impossible to assign the right personnel to the corresponding tasks. This may result in inexperienced personnel handling complex tasks, unable to respond effectively, and may even have a counterproductive effect. At the same time, experienced personnel may be assigned to simpler tasks, resulting in a waste of resources and reduced efficiency.

[0009] 4. No historical events have been preserved

[0010] Problem to be solved: The anti-fraud center has historical records of early warning feedback in anti-fraud scenarios, but the door-to-door dissuasion by the police station is not uniformly retained. As a result, even if there are daily, weekly, and monthly work reports, the details cannot be found. There is no data support for the actual anti-fraud effect, and there are no event processing recordings or video files to prove the processing results.

[0011] 5. Scenario knowledge is not accumulated

[0012] Problem to be solved: In the anti-fraud scenario, the accumulation of relevant knowledge and the lack of learning materials may cause some troubles for staff who have just participated in anti-fraud publicity. At present, the knowledge and information of anti-fraud publicity are usually released through channels such as official accounts, and this information is often not concentrated in one place, which is inconvenient for staff to query and learn. This causes staff to spend a lot of time and energy to find and organize relevant knowledge, affecting work efficiency and quality. In addition, for staff who have just participated in anti-fraud publicity, the lack of relevant learning materials may make them feel at a loss at work. Without systematic learning materials, staff may not be able to fully understand the key content, methods and techniques of anti-fraud publicity, and cannot effectively carry out publicity work. This will not only affect the publicity effect, but may also bring certain risks to anti-fraud work.

[0013] In order to solve the above technical problems, the present invention is proposed. Summary of the invention

[0014] The present invention aims to solve the above problems of the prior art. An anti-fraud volunteer intelligent matching system, method and device are proposed. The technical solution of the present invention is as follows:

[0015] An anti-fraud volunteer intelligent matching system includes: a personnel-event intelligent matching model, a login module, an application scenario management module, a process engine, an event knowledge manual, an element overview module, a personnel management and organization management module, a region management module, an event management module and an equipment management module, wherein the help-seeker login module is used for the help-seeker to log in and send help information to the server, and the server makes a judgment based on the category of the help information and distributes it to volunteers for processing; the application scenario management module is used to define and manage different event scenarios, including naming, describing and setting the scenarios; the process engine is used in different application scenarios, according to the type, level and time multi-dimensional conditions of the event, and combined with the needs of actual business, to output relevant data, including people, events, regions, objects and organizations; the output result data will support the event flow and disposal in the upper-level application; the event knowledge manual module is used to store and manage event-related knowledge and experience, including event processing procedures, common problems and solutions; the element overview module forms a picture of elements through data visualization, It includes people, events, regions, objects and organizations, and displays corresponding element information according to different upper-level applications; the personnel management and organization management modules are used to manage the personnel and organizational structure in the system, including the management of personnel information, authority allocation and organizational structure; the regional management module is used to define and manage the regional scope of the event, including the division, naming and description of the region; the event management module is used to manage the entire process of event creation, allocation, processing and tracking; the equipment management module is used to manage various types of equipment in the system, including equipment registration, configuration and monitoring; the personnel event intelligent matching model includes an event feature clustering module, a processing personnel capability comprehensive evaluation module and an event processing personnel automatic recommendation module, the event feature clustering module includes: event feature extraction step and text vector feature clustering step; the processing personnel capability comprehensive evaluation module includes an indicator design step and a comprehensive evaluation model; when an event occurs, after classifying the event according to the classification, it is recommended based on general principles + special principles, where the special principles do not conflict with each other and the priority of the special principles is greater than the general principles.

[0016] Furthermore, the event feature extraction step specifically includes:

[0017] Forming event feature texts; obtaining a historical event feature text set;

[0018] Extract semantic features of event feature text based on Doc2Vec model and convert text information into real vector information;

[0019] Reduce the dimension of feature vector based on t-SNE method;

[0020] Scale in each dimension according to min-max to (0,1);

[0021] Furthermore, the text vector feature clustering step specifically includes: clustering based on the obtained event feature matrix to achieve classification of historical events; clustering using a density-based Mean Shift clustering method.

[0022] The bandwidth parameter h that affects the model search radius is calculated based on Silverman's rule:

[0023]

[0024] Among them, std(x) represents the standard deviation of the Euclidean distance between two feature vectors, and IQR represents the interquartile range of the Euclidean distance between two feature vectors;

[0025] ④Define the access count matrix A n×n ={a ij}, where a ij Represents v′ i The number of visits to cluster j.

[0026] ⑤Execute Mean Shift process:

[0027] d. Choose any data point v′ k As a starting point;

[0028] The number of visits to cluster k by each point within the bandwidth increases by 1;

[0029] Based on the Euclidean distance, v is calculated within the bandwidth k The density gradient vector grad(v′ k ), which represents the density variation in different directions around the point;

[0030] Move the point a small step in the direction of the maximum density gradient, setting the step size to 0.01;

[0031] The above process is repeated until the moving distance is less than the threshold (0.001);

[0032] e. Record v′ at this time k If the distance between the location and the existing cluster center is greater than the threshold, it is taken as the cluster center;

[0033] f. Repeat the above process a to b until every point is visited;

[0034] ⑥ According to the visit count matrix, classify the points into the class with the most visits.

[0035] Furthermore, the comprehensive evaluation model specifically includes:

[0036] ⑥Indicator data acquisition and cleaning;

[0037] ⑦Weight allocation based on linear regression model:

[0038] c. Establish a linear regression model:

[0039] R=w 0 +w 1 x 1 +w 2 x 2 +w 3 x 3 +w 4 x 4

[0040] Among them, R is the sum of the rankings of various indicators, x 1 ~x 4 Represents 4 indicator values ​​respectively;

[0041] d. Solve the weight W based on the least squares method = (w 0 ,w 1 ,w 2 ,w 3 ,w 4 );

[0042] ⑧ Calculate R': For each event handler, based on the same type of events obtained, collect the values ​​of various indicators X' = (1, x 1 ,x 2 ,x 3 ,x 4 ), calculated according to the weights obtained in ②:

[0043] R′=WX′ T ;

[0044] ⑨ Scaling R′: Scale the possible R cases in (0,0.6) according to min-max, where: min = 4*1, min case means that the four indicators are all ranked first, max = 4*N, N is the number of participants in the evaluation, and max case means that the four indicators are all ranked last, and get R′ 0 ;

[0045] ⑩Calculate the comprehensive score S = (1-R′ 0 )*100%.

[0046] An anti-fraud volunteer intelligent matching method based on any of the above systems, comprising the following steps:

[0047] The server receives the help-seeking information sent by the help-seeker via the NB-IoT-based wireless communication module; determines the type of volunteer needed by the help-seeking information based on the personnel-event intelligent matching model; matches the anti-fraud volunteers in the volunteer information database based on the pre-set volunteer discrimination and classification rules and the volunteer category in combination with the event knowledge manual, and obtains the matched anti-fraud volunteers based on the personnel management and organization management modules; and forwards the help-seeking information to the matched volunteer's terminal for processing based on the regional management module.

[0048] Furthermore, before the server receives the help information sent by the help seeker via the NB-IoT-based wireless communication module, it also includes: receiving the check-in information sent by the volunteer's terminal via the NB-IoT-based wireless communication module; and organizing the information of the volunteer who sent the check-in information into the volunteer information database.

[0049] Furthermore, the NB-IoT-based wireless communication module is a NB-IoT-based wireless communication module that reads or receives information in the terminal of the person seeking help near the person seeking help, and emits a prompt tone when the NB-IoT-based wireless communication module receives the help information.

[0050] Furthermore, the step of forwarding the help-seeking information to the terminal of the matched volunteer includes: determining whether the anti-fraud volunteer is in the scenic area where the help-seeker is located; if not, notifying the anti-fraud volunteer through the wireless communication network; if yes, notifying the anti-fraud volunteer through the nearest NB-IoT.

[0051] Furthermore, matching volunteers in the volunteer information database according to the preset rules and the volunteer categories includes:

[0052] According to the category of anti-fraud volunteers, qualified anti-fraud volunteers are screened out in the volunteer information database; among the qualified anti-fraud volunteers, it is searched whether there are volunteers within the set range, and if so, the anti-fraud volunteer closest to the person seeking help is selected as the matching object; if not, volunteers of a category similar to that of the anti-fraud volunteers are screened out in the volunteer information database, and the volunteer closest to the person seeking help within the set range is selected as the matching object.

[0053] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the anti-fraud volunteer intelligent matching method as described in any one of the items is implemented.

[0054] The advantages and beneficial effects of the present invention are as follows:

[0055] The present invention realizes the full-process digital management of data and events in the process of urban governance, and at the same time, it precipitates the business knowledge and business processes of public security-related scenarios into the knowledge base, and retains, analyzes and summarizes the business results, so as to reduce the burden on grassroots personnel in town and street governance, thereby achieving the best governance effect. The present invention can improve the accuracy of volunteer matching and improve the management efficiency of personnel through the wireless communication module based on NB-IoT combined with the anti-fraud volunteer intelligent matching system. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a block diagram of the anti-fraud volunteer intelligent matching system provided by the present invention in a preferred embodiment;

[0057] Figure 2 It is a personnel management diagram;

[0058] Figure 3 It is a schematic diagram of anti-fraud process management;

[0059] Figure 4 It is a personnel management diagram;

[0060] Figure 5 It is a schematic diagram of historical scene query;

[0061] Figure 6 A schematic diagram showing the knowledge base used to accumulate anti-fraud knowledge. DETAILED DESCRIPTION

[0062] The following will describe the technical solutions in the embodiments of the present invention in detail in conjunction with the accompanying drawings in the embodiments of the present invention. The described embodiments are only part of the embodiments of the present invention.

[0063] The technical solution of the present invention to solve the above technical problems is: Figure 1As shown, an anti-fraud volunteer intelligent matching system includes: a personnel event intelligent matching model, a login module, an application scenario management module, a process engine, an event knowledge manual, an element overview module, a personnel management and organization management module, a regional management module, an event management module and an equipment management module, wherein the help-seeker login module is used for the help-seeker to log in and send help information to the server, and the server makes a judgment based on the category of the help information and distributes it to volunteers for processing; the application scenario management module is used to define and manage different event scenarios, including naming, describing and setting the scenarios; the process engine is used in different application scenarios, according to the type, level and time of the event multi-dimensional conditions, and combined with the actual business needs, to output relevant data, including people, events, regions, objects and organizations; the output result data will support the event flow and disposal in the upper-level application; the event knowledge manual module is used to store and manage event-related knowledge and experience, including event processing procedures, common problems and solutions; the element overview module forms a picture of elements through data visualization. , which includes people, events, regions, objects and organizations, and displays corresponding element information according to different upper-level applications; the personnel management and organization management modules are used to manage the personnel and organizational structure in the system, including the management of personnel information, authority allocation and organizational structure; the regional management module is used to define and manage the regional scope of the event, including the division, naming and description of the region; the event management module is used to manage the entire process of event creation, allocation, processing and tracking; the equipment management module is used to manage various types of equipment in the system, including equipment registration, configuration and monitoring; the personnel event intelligent matching model includes an event feature clustering module, a processing personnel capability comprehensive evaluation module and an event processing personnel automatic recommendation module, the event feature clustering module includes: event feature extraction step and text vector feature clustering step; the processing personnel capability comprehensive evaluation module includes an indicator design step and a comprehensive evaluation model; when an event occurs, after the event is classified according to the classification, it is recommended based on general principles + special principles, where the special principles do not conflict with each other and the priority of the special principles is greater than the general principles.

[0064] Furthermore, the event feature extraction step specifically includes:

[0065] Forming event feature texts; obtaining a historical event feature text set;

[0066] Extract semantic features of event feature text based on Doc2Vec model and convert text information into real vector information;

[0067] Reduce the dimension of feature vector based on t-SNE method;

[0068] Scale in each dimension according to min-max to (0,1);

[0069] Furthermore, the text vector feature clustering step specifically includes: clustering based on the obtained event feature matrix to achieve classification of historical events; clustering using a density-based Mean Shift clustering method.

[0070] The bandwidth parameter h that affects the model search radius is calculated based on Silverman's rule:

[0071]

[0072] Among them, std(x) represents the standard deviation of the Euclidean distance between two feature vectors, and IQR represents the interquartile range of the Euclidean distance between two feature vectors;

[0073] ⑦Define the access count matrix A n×n ={a ij}, where a ij Represents v′ i The number of visits to cluster j.

[0074] ⑧Execute Mean Shift process:

[0075] g. Choose any data point v′ k As a starting point;

[0076] The number of visits to cluster k by each point within the bandwidth increases by 1;

[0077] Based on the Euclidean distance, v is calculated within the bandwidth k The density gradient vector grad(v′ k ), which represents the density variation in different directions around the point;

[0078] Move the point a small step in the direction of the maximum density gradient, setting the step size to 0.01;

[0079] The above process is repeated until the moving distance is less than the threshold (0.001);

[0080] h. Record v′ at this time k If the distance between the location and the existing cluster center is greater than the threshold, it is taken as the cluster center;

[0081] i. Repeat the above process a to b until every point is visited;

[0082] ⑨ According to the visit count matrix, classify the points into the class with the most visits.

[0083] Furthermore, the comprehensive evaluation model specifically includes:

[0084] Acquisition and cleaning of indicator data;

[0085] Weight allocation based on linear regression model:

[0086] e. Establish a linear regression model:

[0087] R=w 0 +w 1 x 1 +w 2 x 2 +w 3 x 3 +w 4 x 4

[0088] Among them, R is the sum of the rankings of various indicators, x 1 ~x 4 Represents 4 indicator values ​​respectively;

[0089] f. Solve the weight W based on the least squares method = (w 0 ,w 1 ,w 2 ,w 3 ,w 4 );

[0090] Calculate R': For each event handler, based on the same type of events obtained, collect the values ​​of various indicators X' = (1, x 1 ,x 2 ,x 3 ,x 4 ), calculated according to the weights obtained in ②:

[0091] R′=WX′ T ;

[0092] Scaling R′: Scale the possible R cases in (0,0.6) according to min-max, where: min = 4*1, min case means that the four indicators are all ranked first, max = 4*N, N is the number of participants in the evaluation, and max case means that the four indicators are all ranked last, and get R′ 0 ;

[0093] Calculate the comprehensive score S = (1-R' 0 )*100%.

[0094] An anti-fraud volunteer intelligent matching method based on any of the above systems, comprising the following steps:

[0095] The server receives the help-seeking information sent by the help-seeker via the NB-IoT-based wireless communication module; determines the type of volunteer needed by the help-seeking information based on the personnel-event intelligent matching model; matches the anti-fraud volunteers in the volunteer information database based on the pre-set volunteer discrimination and classification rules and the volunteer category in combination with the event knowledge manual, and obtains the matched anti-fraud volunteers based on the personnel management and organization management modules; and forwards the help-seeking information to the matched volunteer's terminal for processing based on the regional management module.

[0096] To solve problem 1, Figure 2 As shown, the solution / idea is: the knowledge base manages the staff in a unified manner, and designs corresponding roles for each scenario according to the scenario business processing flow. Each role has different permissions, and then assigns personnel to each role group so that the work can be reasonably distributed to improve the efficiency and effectiveness of scenario business processing. Using the permission rule module, in the upper and lower management, the upper department has higher permissions to view the data of the lower department.

[0097] To solve problem 2, Figure 3 As shown, the solution / idea is: the knowledge base will sort out and manage the door-to-door dissuasion process of anti-fraud scenarios, clarify each event type, and the specific dissuasion process in different scenarios. For example, for student-related events, teachers in the school can be given priority to dissuasion, and community grid workers can be given priority to dissuasion for older people. Volunteers can be given priority to dissuasion for white-collar workers in the company. In addition, for events with large amounts or multiple warnings, police officers are required to dissuasion. This can clearly define the entire process and provide a basis for subsequent management and optimization. The door-to-door dissuasion process is visualized and editable through the process engine. The process engine can display the process graphically, allowing users to clearly see the order and relationship of each step and link. At the same time, users can also edit through the process engine to adjust and optimize the process according to actual needs.

[0098] To solve problem 3, Figure 4As shown, the solution / idea: The knowledge base can collect and sort out event information in the anti-fraud scenario, including victim information, fraud type, region, time, etc. This information can be used for subsequent analysis and screening. Based on the collected event information, the knowledge base can analyze the anti-fraud type. By analyzing different fraud types, we can understand the characteristics, common methods and response strategies of each type of fraud. This helps to provide a basis for the screening of door-to-door dissuasion personnel. Based on the analysis results of event information and anti-fraud types, the knowledge base can screen door-to-door dissuasion personnel. According to the characteristics and requirements of different anti-fraud types, a series of screening conditions can be set, such as experience, professional background, region, etc. Through screening, a list of optimal processing personnel who meet the conditions can be obtained. The knowledge base can provide a list of optimal processing personnel based on the screening results of door-to-door dissuasion personnel. This list can be selected by the dissuasion task issuer to ensure that each task can be assigned to the most suitable person for processing. The task issuer can refer to the list of optimal processing personnel based on the characteristics and requirements of the task and select the most suitable person to assign the dissuasion task.

[0099] To solve problem 4, Figure 5 As shown, the solution / idea: The knowledge base system can not only precipitate anti-fraud knowledge, but also effectively manage the feedback and data of anti-fraud events. In the whole process of dissuasion events, the basic information of the event, the door-to-door dissuasion information, including the information of the parties, the video recorded or taken by the dissuasion personnel, and the feedback information of the final dissuasion will be synchronized to the knowledge base. Therefore, the knowledge base also has the function of statistical analysis and query of historical events. Through the statistical analysis of historical events, we can understand the overall situation of dissuasion work, including the number of dissuasion, success rate, frequency of fraud methods, etc. These data can help the anti-fraud department evaluate the effectiveness and problems of anti-fraud work, and provide data support for the formulation of more effective anti-fraud strategies. At the same time, the knowledge base also provides the function of querying historical event details, and staff can find specific events and related information as needed.

[0100] To solve problem 5, Figure 6As shown, the solution / idea: The knowledge base can be used to precipitate anti-fraud knowledge, form an anti-fraud knowledge base, and improve the success rate of dissuasion and the effectiveness of anti-fraud publicity. In the knowledge base, publicity and guidance are carried out through two sections. In the knowledge publicity section, common fraud methods, real cases and police anti-fraud tips are displayed, and various common fraud methods are listed, such as telephone fraud, Internet fraud, SMS fraud, etc., and the characteristics and common methods of each method are introduced in detail. For example, in telephone fraud, fraudsters often impersonate public security, bank staff and other identities, and use threats, intimidation, deception and other means to defraud victims of their money. At the same time, door-to-door dissuasion personnel can view specific dissuasion scripts for the current incident and provide targeted guidance to door-to-door dissuasion personnel. The knowledge base will form a comprehensive anti-fraud knowledge base, providing dissuasion personnel with rich information and tools to help them better carry out anti-fraud work.

[0101] Preferably, it also includes an anti-fraud volunteer intelligent matching method, which includes the following steps:

[0102] The server receives the help-seeking information sent by the help-seeker via the NB-IoT-based wireless communication module; according to the application scenario management module, determines the category of volunteers needed by the help-seeking person based on the help-seeking information; in combination with the event knowledge manual, matches the anti-fraud volunteers in the volunteer information database according to the pre-set volunteer discrimination and classification rules and the volunteer category, and obtains the matched anti-fraud volunteers according to the personnel management and organization management modules; and forwards the help-seeking information to the matched volunteer's terminal for processing according to the regional management module.

[0103] Preferably, before the server receives the help information sent by the help seeker via the wireless communication module based on NB-IoT, it also includes: receiving the check-in information sent by the volunteer's terminal via the wireless communication module based on NB-IoT; and organizing the information of the volunteer who sent the check-in information into the volunteer information database.

[0104] Preferably, the NB-IoT-based wireless communication module is a NB-IoT-based wireless communication module that reads or receives information in the terminal of the person seeking help near the person seeking help, and emits a prompt tone when the NB-IoT-based wireless communication module receives the help information.

[0105] Preferably, the step of forwarding the help-seeking information to the terminal of the matched volunteer includes: determining whether the anti-fraud volunteer is in the scenic area where the help-seeker is located; if not, notifying the anti-fraud volunteer through the wireless communication network; if yes, notifying the anti-fraud volunteer through the nearest NB-IoT.

[0106] Preferably, matching volunteers in the volunteer information database according to preset rules and the volunteer category includes: screening out qualified anti-fraud volunteers in the volunteer information database according to the anti-fraud volunteer category; searching among the qualified anti-fraud volunteers whether there are volunteers within a set range, and if so, selecting the anti-fraud volunteer closest to the helper as a matching object; if not, screening out volunteers of a category similar to the anti-fraud volunteer category in the volunteer information database, and selecting the volunteer closest to the helper within the set range as a matching object.

[0107] Preferably, the personnel event intelligent matching model of the present invention is:

[0108] Event feature clustering

[0109] Event feature extraction: When the platform receives an event processing request, it contains event feature information including description content, event type, subcategory name, occurrence time, town / street, etc. Based on this data, we convert the text into a multi-dimensional event feature vector.

[0110] 1) For each event, the event type, subcategory name, occurrence time, and town / street to which it belongs are modified into a sentence structure of "event of" + "event type / subcategory name / occurrence time / town / street to which it belongs" + "is XXXX", and concatenated with the event content description with commas to form event feature text. The historical event feature text set W = {w 1 ,w 2 ,…,w n}, where w i The event feature text representing the i-th event;

[0111] 2) Extract semantic features of event feature text based on Doc2Vec model and convert text information into real vector information. Doc2Vec is a text embedding model based on neural network, which is used to map documents into continuous vector space to capture semantic relationships between documents. Doc2Vec calculation can convert both vocabulary and documents into document vectors in real domain and retain semantic similarity in vector space. After model calculation, we get event feature matrix V n×200 =(v 1 ,v 2 ,…,v n ) T ,i=1,2,…,n, where v i is the document vector of the i-th event, with a dimension of 200;

[0112] 3) Reduce the dimension of the feature vector based on the t-SNE method. The document vector obtained in 2) has a high dimension, which will cause the data points represented by the vector to be sparsely distributed in the high-dimensional space, affecting the subsequent clustering effect. Therefore, we use the t-SNE method to reduce the dimension of the document vector and obtain the event feature matrix V′ after dimensionality reduction. n×6 =(v′ 1 ,v′ 2 ,…,v′ n ) T ,i=1,2,…,n;

[0113] 4) Scale the vectors in V′ according to min-max in each dimension to (0,1);

[0114] Text vector feature clustering

[0115] In this part, we need to cluster based on the event feature matrix obtained in 3.1.1 to achieve the classification of historical events. Because we cannot know the exact number of clusters and the dimension of the feature vector is still high, we use the density-based Mean Shift clustering method for clustering.

[0116] ⑩Calculate the bandwidth parameter h that affects the model search radius based on Silverman's rule:

[0117]

[0118] Among them, std(x) represents the standard deviation of the Euclidean distance between two feature vectors, and IQR represents the interquartile range of the Euclidean distance between two feature vectors;

[0119] Define the access count matrix A n×n ={a ij}, where a ij Represents v′ i The number of visits to cluster j.

[0120] Execute the Mean Shift process:

[0121] j. Choose any data point v′ k As a starting point;

[0122] The number of visits to cluster k by each point within the bandwidth increases by 1;

[0123] Based on the Euclidean distance, v is calculated within the bandwidth k The density gradient vector grad(v′ k ), which represents the density variation in different directions around the point;

[0124] Move the point a small step in the direction of the maximum density gradient (set the step size to 0.01);

[0125] The above process is repeated until the moving distance is less than the threshold (0.001);

[0126] k. Record v′ at this time k If the distance between the location and the existing cluster center is greater than the threshold, it is taken as the cluster center;

[0127] l. Repeat the above process a~b until every point is visited;

[0128] According to the visit count matrix, the points are classified into the class with the most visits;

[0129] Comprehensive evaluation of personnel capabilities

[0130] Indicator design

[0131] In order to evaluate the ability of personnel to handle incidents, we set up four indicators based on system monitoring data:

[0132] Average total time spent on incident handling: the average time spent by personnel on handling this type of incident from the time they arrive at the scene to the time they complete the incident handling for nearly 50 incidents, in hours;

[0133] Arrival time for incident handling: the average time for personnel to receive incident handling assignments and arrive at the site to handle the nearly 50 incidents, in hours;

[0134] Client satisfaction rate: the proportion of cases in which the parties were satisfied with the results of follow-up visits among the nearly 50 cases handled;

[0135] Success rate of handling: the proportion of cases in which the parties concerned have resolved the issues and have not submitted another handling request for the same case among the nearly 50 cases handled;

[0136] Comprehensive evaluation model

[0137] Taking the above indicators into consideration, calculate the comprehensive evaluation score of the processing personnel according to the following steps:

[0138] Acquisition and cleaning of indicator data;

[0139] Weight allocation based on linear regression model:

[0140] g. Establish a linear regression model:

[0141] R=w 0 +w 1 x 1 +w2 x 2 +w 3 x 3 +w 4 x 4

[0142] Among them, R is the sum of the rankings of various indicators, x 1 ~x 4 Respectively represent the 4 indicator values ​​in 3.2.1;

[0143] h. Solve the weight W based on the least squares method = (w 0 ,w 1 ,w 2 ,w 3 ,w 4 );

[0144] Calculate R': For each event handler, based on the similar events obtained in 3.1, collect the index values ​​X'=(1,x 1 ,x 2 ,x 3 ,x 4 ), calculated according to the weights obtained in ②:

[0145] R′=WX′ T ;

[0146] Scaling R′: Scale the possible R cases in (0,0.6) according to min-max, where: min = 4*1 (min case means that the four indicators are all ranked first), max = 4*N (N is the number of participants in the evaluation, and the max case means that the four indicators are all ranked last), and get R′ 0 ;

[0147] Calculate the comprehensive score S = (1-R' 0 )*100%

[0148] Automatic recommendation of incident handlers

[0149] When an event occurs, after classifying the event according to the classification in 3.1, recommendations are made based on general principles + special principles, where special principles do not conflict with each other and special principles have a higher priority than general principles.

[0150] General principles: Those with high scores in the history of handling such incidents will be given priority;

[0151] Special principles: When multiple events occur simultaneously, the recommendation will be made based on the principle of the highest total score of the person handling the event;

[0152] When the scores of all available personnel are lower than the threshold (70 points by default), priority will be given to those who have not handled such incidents (or have handled them 5 times or less). The order of recommendation is from high to low based on the scores of all incidents handled.

[0153] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the anti-fraud volunteer intelligent matching method as described in any one of the items is implemented.

[0154] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0155] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0156] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0157] The above embodiments should be understood to be only used to illustrate the present invention and not to limit the protection scope of the present invention. After reading the contents of the present invention, technicians can make various changes or modifications to the present invention, and these equivalent changes and modifications also fall within the scope defined by the claims of the present invention.

Claims

1. An anti-fraud volunteer intelligent matching system, characterized in that: include: Personnel-event intelligent matching model, login module, application scenario management module, process engine, event knowledge manual, element overview module, personnel management and organization management module, area management module, event management module and equipment management module. Among them, the help-seeker login module is used for help-seekers to log in and send help information to the server. The server determines the type of help information and distributes it to volunteers for processing; the application scenario management module is used to define and manage different event scenarios, including naming, describing and setting the scenarios; the process engine is used in different application scenarios to output relevant data, including people, events, areas, objects and organizations according to the multi-dimensional conditions of event type, level and time, and combined with actual business needs; the output result data will support the event flow and disposal in the upper-level application; the event knowledge manual module is used to store and manage event-related knowledge and experience, including event processing procedures, common problems and solutions; the element overview module forms an element picture through data visualization, which includes people, events, Regions, objects and organizations, and display corresponding element information according to different upper-level applications; personnel management and organization management modules are used to manage personnel and organizational structures in the system, including the management of personnel information, authority allocation and organizational structure; regional management module is used to define and manage the regional scope of events, including the division, naming and description of regions; event management module is used to manage the entire process of event creation, allocation, processing and tracking; equipment management module is used to manage various types of equipment in the system, including equipment registration, configuration and monitoring; personnel event intelligent matching model, including event feature clustering module, processing personnel capability comprehensive evaluation module and event processing personnel automatic recommendation module, the event feature clustering module includes: event feature extraction step and text vector feature clustering step; the processing personnel capability comprehensive evaluation module includes indicator design step and comprehensive evaluation model; when an event occurs, after classifying the event according to the classification, it is recommended based on general principles + special principles, where special principles do not conflict with each other and special principles have higher priority than general principles; The event feature extraction step specifically includes: Forming event feature texts; obtaining a historical event feature text set; Extract semantic features of event feature text based on Doc2Vec model and convert text information into real vector information; Reduce the dimension of feature vector based on t-SNE method; Scale in each dimension according to min-max to (0,1); The text vector feature clustering step specifically includes: clustering based on the obtained event feature matrix to achieve classification of historical events; clustering using the density-based Mean Shift clustering method; The bandwidth parameter h that affects the model search radius is calculated based on Silverman's rule: ； in, represents the standard deviation of the Euclidean distance between two eigenvectors, The interquartile range represents the Euclidean distance between two eigenvectors; ①Define the visit count matrix ,in express The number of visits to cluster j; ②Execute the Mean Shift process: a. Choose any data point As a starting point; The number of visits to cluster k by each point within the bandwidth increases by 1; Based on Euclidean distance, calculated within the bandwidth The density gradient vector , represents the density variation in different directions around the point; Move the point a small step in the direction of the maximum density gradient, setting the step size to 0.01; The above process is repeated until the moving distance is less than the threshold value 0.001; b. Record this time If the distance between the location and the existing cluster center is greater than the threshold, it is taken as the cluster center; c. Repeat the above process a~b until every point is visited; ③ According to the access times matrix, classify the points into the class with the most access times; The comprehensive evaluation model specifically includes: S1: indicator data acquisition and cleaning; S2: Weight allocation based on linear regression model: S201: Establish linear regression model: Among them, R is the sum of the rankings of various indicators. Represents 4 indicator values ​​respectively; S202: Solving weights based on the least squares method ; S3: Compute :For each event handler, based on similar events obtained, collect various indicator values , calculated according to the weights obtained in S2: ； S4: Yes Scaling: Scale the possible R cases in (0,0.6) according to min-max, where min=4*1, min means the candidate is ranked first in all four indicators, max=4*N, N is the number of participants in the scoring, max means the candidate is ranked last in all four indicators, and we get ; S5: Calculate the overall score .

2. An anti-fraud volunteer intelligent matching method based on the system described in claim 1, characterized in that: The steps include: The server receives the help information sent by the help seeker via the NB-IoT-based wireless communication module; According to the personnel-event intelligent matching model, the type of volunteer needed by the help-seeker is determined according to the help-seeking information; in combination with the event knowledge manual, the anti-fraud volunteers are matched in the volunteer information database according to the pre-set volunteer discrimination and classification rules and the volunteer categories, and the matched anti-fraud volunteers are obtained according to the personnel management and organization management modules; The help information is forwarded to the terminal of the matched volunteer for processing according to the regional management module.

3. According to claim 2, an anti-fraud volunteer intelligent matching method is characterized in that: Before the server receives the help information sent by the help seeker via the NB-IoT based wireless communication module, it also includes: receiving the check-in information sent by the volunteer's terminal via the NB-IoT based wireless communication module; and organizing the information of the volunteer who sent the check-in information into the volunteer information database.

4. According to claim 3, the method for intelligent matching of anti-fraud volunteers is characterized in that: The NB-IoT-based wireless communication module is a NB-IoT-based wireless communication module that reads or receives information in the terminal of the person seeking help near the person seeking help. When the NB-IoT-based wireless communication module receives the help information, a prompt tone is issued.

5. The anti-fraud volunteer intelligent matching method according to claim 4 is characterized in that: The step of forwarding the help-seeking information to the terminal of the matched volunteer includes: determining whether the anti-fraud volunteer is in the scenic area where the help-seeker is located; if not, notifying the anti-fraud volunteer through the wireless communication network; if yes, notifying the anti-fraud volunteer through the nearest NB-IoT.

6. The method for intelligent matching of anti-fraud volunteers according to claim 2 is characterized in that: The matching of volunteers in the volunteer information database according to the preset rules and the volunteer categories includes: According to the category of anti-fraud volunteers, qualified anti-fraud volunteers are screened out in the volunteer information database; among the qualified anti-fraud volunteers, it is searched whether there are volunteers within the set range, and if so, the anti-fraud volunteer closest to the person seeking help is selected as the matching object; if not, volunteers of a category similar to that of the anti-fraud volunteers are screened out in the volunteer information database, and the volunteer closest to the person seeking help within the set range is selected as the matching object.

7. An electronic device, characterized in that: It includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the program, it implements the anti-fraud volunteer intelligent matching method as described in any one of claims 2 to 6.

Images