Personnel feature series-parallel method based on organizational structure diagram

Abstract

The invention discloses a personnel feature series-parallel method based on an organizational structure diagram. Historical related personnel elements are input for sample training; analyzing and generating key element features; forming a key element and related object decision tree; inputting a new related object condition set; performing intelligent mining analysis based on a decision tree; and the abnormal object is accurately pushed. According to the personnel feature series-parallel method based on the organizational structure diagram, based on new technologies such as cloud computing and big data, an existing big data resource system is fully utilized, through algorithms and means such as data mining and machine learning, an intelligent mining analysis model based on historical cases is constructed, the latest machine learning and deep learning algorithms are integrated, and the personnel feature series-parallel method based on the organizational structure diagram is obtained. The method improves the accuracy of related behavior recognition, provides an intelligent application handling scene for a worker through automatic learning of historical case investigation experience, intelligent analysis and inference, and precise pushing of an abnormal object of a new event, and achieves a new situation of intelligent handling and intelligent application handling.

Description

Technical Field

[0001] The present invention relates to the field of fraud investigation and analysis application technology, and more specifically to a personnel feature stringing and paralleling method based on an organizational chart. Background Art

[0002] The method of combining and combining personnel characteristics has important application value in the field of fraud investigation and judgment. With the development of technology, a variety of solutions have appeared on the market, but they still have shortcomings in some aspects. Many grassroots department staff who have no experience in handling cases face a variety of people every day in the process of handling related events. Clues are scarce and the case is complicated. They often have no idea where to start. There are mainly the following situations:

[0003] First, the shortcomings of existing solutions: existing technologies may not be able to accurately identify the patterns of fraudulent behavior, resulting in false positives or omissions. Fraudulent methods are constantly evolving. Existing systems may lack the ability to update and learn new fraud patterns in real time. Solutions may be too general and unable to meet the specific needs of different cases and different staff members. Second, technical or cognitive reasons: Technology is developing rapidly, but existing solutions may not keep up with the latest technological trends. Existing systems may not fully utilize data science and artificial intelligence technologies to improve analytical capabilities. The operating habits and case handling procedures of grassroots staff may not be fully considered during design. The system may lack the ability to self-learn and adapt to new fraudulent methods. Based on the above reasons, it is difficult to combine the characteristics of the current case (such as approach methods, infringement methods, object selection, timing selection, area selection, etc.) and clue elements (time, space, personnel, objects, etc.) to achieve accurate analysis and judgment, and automatically analyze potential suspicious objects.

[0004] In order to solve this problem, the current traditional approach is to target the characteristics and patterns related to personal fraud, set certain association rules based on the relationship between the characteristics of the crime and the suspicious objects, and infer the most likely potential abnormal objects through comprehensive calculations of the association rules.

[0005] Although this technical means has discovered potential targets of crime to a certain extent and partially solved the problems of automated calculations and automatic push of potential targets, the design idea of ​​this solution does not look for patterns in historical fraud samples, but only relies on the experience of case experts to summarize, provide relevant association rules, and set scoring weights. It has certain drawbacks: first, it is difficult to achieve accurate reasoning calculations, and second, it cannot adapt to the analysis of new fraud methods. The overall effectiveness is greatly reduced and cannot meet the requirements of practical applications. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a personnel feature string and parallel method based on the organizational structure chart, which can automatically analyze potential abnormal objects, accurately push suspicious objects of new cases, and provide intelligent processing application scenarios for staff.

[0007] In order to solve the above technical problems, the technical solutions adopted by the present invention are as follows.

[0008] The personnel characteristics concatenation method based on the organizational chart includes the following steps:

[0009] A. Select typical cases from the historical analysis and judgment factors of relevant personnel, and input historical relevant personnel factors for sample training;

[0010] B. Analyze and generate key element characteristics;

[0011] C. Form a decision tree of key elements and related objects;

[0012] D. Enter a new set of related object conditions;

[0013] E. Intelligent mining and analysis based on decision tree;

[0014] F. Accurately push abnormal objects.

[0015] To further optimize the technical solution, in step A, sample training includes data collection and information integration, data collection includes organizational structure information and basic personnel information; information integration includes multi-source data integration and information cleaning.

[0016] To further optimize the technical solution, in step B, the key elements of analysis include data preprocessing, key field recognition, fuzzy matching technology and OCR recognition.

[0017] To further optimize the technical solution, in step C, the decision tree of key elements and related objects includes intelligent serialization and parallelization, rule engine, machine learning algorithm, relationship network construction, intersection and correlation point analysis, and search and indexing technology.

[0018] To further optimize the technical solution, in step D, setting the object condition set includes data reading, role comparison and organizational chart drawing.

[0019] To further optimize the technical solution, in step E, the intelligent mining analysis of the decision tree includes correlation calculation, machine learning algorithm, big data technology, cloud computing platform, intelligent and precise serial and parallel, data mining algorithm integration and key information matching.

[0020] To further optimize the technical solution, in step F, accurate push of abnormal objects includes serial-parallel generation, serial-parallel push, push and feedback, visual display, real-time update and dynamic matching, and reminder mechanism.

[0021] Due to the adoption of the above technical scheme, the technical progress achieved by the present invention is as follows.

[0022] The personnel feature stringing and paralleling method based on the organizational structure chart provided by the present invention is based on new technologies such as cloud computing and big data, makes full use of the existing big data resource system, and constructs an intelligent mining and analysis model based on historical cases through algorithms and means such as data mining and machine learning. It integrates the latest machine learning and deep learning algorithms to improve the accuracy of related behavior recognition, and accurately pushes abnormal suspicious objects of new events through automatic learning of historical case investigation experience, intelligent analysis and inference, and precise push of intelligent handling application scenarios for staff, thus realizing a new situation of smart handling and intelligent handling applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0024] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0025] Based on the personnel characteristics serial and parallel method of the organizational structure chart, combined with Figure 1 As shown, it includes the following steps:

[0026] A. Select typical examples from the historical analysis and judgment factors of relevant personnel, and input historical relevant personnel factors for sample training.

[0027] Sample training includes data collection and information integration. Data collection includes organizational structure information and basic personnel information. Organizational structure information is used to collect detailed structural information of the target organization, including levels, departments, teams, etc.; basic personnel information is used to collect basic information of personnel, such as name, position, nickname, contact information, etc.

[0028] Information integration includes multi-source data integration and information cleaning. Multi-source data integration is to integrate personnel information from different sources; information cleaning is to clean the collected data to ensure the accuracy and consistency of the data.

[0029] B. Analyze and generate key element characteristics.

[0030] Use statistical methods or machine learning algorithms to mark anomalies. According to the possibility of different combinations of elements and behavioral characteristics of historical suspects, set the support and confidence of each combination, find some frequent item sets, and connect the frequent item sets to find the frequent item sets with the largest number of factors. The factors in the frequent item sets are the key items to be selected as the potential suspicious objects in the case.

[0031] The key elements of the analysis include data preprocessing, key field identification, fuzzy matching technology and OCR recognition.

[0032] Data preprocessing includes data cleaning and data standardization. Data cleaning is used to remove noise and inconsistencies in the data, such as format errors, duplicate records, etc.; data standardization is to convert data in different formats into a unified format for subsequent processing.

[0033] Key field identification includes field selection and field weight. Field selection is to determine the key fields used for matching, such as name, nickname, ID number, den name, company name, etc.; field weight is to assign weights to different fields to reflect their importance in the matching process.

[0034] Fuzzy matching technology includes string similarity calculation and fuzzy query. String similarity calculation uses algorithms such as Levenshtein distance and Jaro-Winkler method to calculate the similarity between strings; fuzzy query allows a certain degree of difference when matching fields, thereby improving the flexibility of matching.

[0035] OCR recognition includes image recognition and data conversion. Data recognition uses OCR technology to recognize unstructured text, such as text in scanned documents or pictures; data conversion converts the recognized text into analyzable structured data.

[0036] C. Form a decision tree of key elements and related objects.

[0037] After selecting the decisive factor feature items, the mapping relationship between the feature item attributes and the relevant object types is set, and decisions are made on new cases through the decision tree algorithm. Each fork path represents a possible factor feature value, and each leaf node corresponds to the object represented by the path from the root node to the leaf node. By comprehensively analyzing historical case data and relevant personnel characteristics, and using data mining and machine learning techniques, the efficiency and accuracy of related case investigations can be effectively improved.

[0038] The decision tree of key elements and related objects includes intelligent concatenation and parallelization, rule engine, machine learning algorithm, relationship network construction, intersection and correlation point analysis, and search and indexing technology.

[0039] Intelligent concatenation includes key field matching and association analysis. Key field matching is to achieve intelligent matching based on key fields such as flower names, den names, company names, and personnel roles; association analysis is to use data mining technology to identify and associate personnel relationships within an organization or between multiple organizations.

[0040] The rule engine includes rule definition and rule application. Rule definition is to define matching rules according to business needs. For example, people with the same name and surname and in the same den may belong to the same gang. Rule application is to apply these rules in the data matching process to identify potential associations.

[0041] Machine learning algorithms include data mining, machine learning frameworks, and cluster analysis. Data mining is used to discover patterns and associations in data; machine learning frameworks such as scikit-learn, TensorFlow, or PyTorch are used to achieve more advanced data analysis and prediction; cluster analysis is the application of clustering algorithms, such as K-means or DBSCAN, to group similar records.

[0042] The relational network is constructed as a graph database, which stores information about people and organizations in the graph database to facilitate network analysis.

[0043] Intersection and correlation point analysis includes intersection identification and correlation point mining. Intersection identification is to find the key fields that appear in multiple data sets, find the intersection of personnel between different dens or companies, and identify the intersection; correlation point mining is to identify potential correlation points by analyzing the social relationships, economic exchanges, etc. of personnel.

[0044] Search and indexing technologies include search engines and indexing services. Search engines, such as Elasticsearch, are used to quickly retrieve information in an organizational structure; indexing services are used to improve the efficiency and performance of data queries.

[0045] D. Enter a new set of conditions for fraud-related objects.

[0046] Setting object condition sets includes data reading, role comparison, and organizational chart drawing.

[0047] Data reading includes multi-source data access and data integration, both of which provide data input for new object condition sets.

[0048] Role comparison includes key field matching and rule engine. Key field matching is to match roles based on key fields such as name, position, nickname, etc. The rule engine applies business rules to compare roles, identify the roles of people in different organizations, and is used to identify objects in the new input condition set.

[0049] The drawing of an organizational chart includes graphic representation and hierarchical relationships. Graphical representation is to use software tools to draw an organizational chart to show the relationships between personnel and the organizational structure; hierarchical relationships are to clearly indicate different levels and superior-subordinate relationships, highlighting key figures and groups.

[0050] E. Intelligent mining and analysis based on decision tree.

[0051] Decision tree intelligent mining analysis includes correlation calculation, machine learning algorithm, big data technology, cloud computing platform, intelligent and precise serial and parallel, reagent library mining algorithm integration and key information matching.

[0052] Association calculation includes similarity calculation and association rule mining. Similarity calculation is to calculate the similarity between different data records to determine whether they point to the same entity; association rule mining: using data mining technology to discover potential association rules between data.

[0053] Machine learning algorithms include data mining, machine learning frameworks, and cluster analysis. Data mining is used to discover patterns and associations in data; machine learning frameworks such as scikit-learn, TensorFlow, or PyTorch are used to achieve more advanced data analysis and prediction; cluster analysis is the application of clustering algorithms, such as K-means or DBSCAN, to group similar records.

[0054] Big data technology includes data storage and distributed computing. Data storage is the use of big data technology to store and manage massive data; distributed computing is the use of distributed computing capabilities of cloud computing to process large-scale data sets.

[0055] The cloud computing platform includes elastic resources and service reliability. Elastic resources dynamically adjust resources according to system load to ensure high performance; service reliability utilizes the high availability and disaster recovery capabilities of cloud services.

[0056] Intelligent and precise serial and parallel processing includes machine learning and model training. Machine learning is the application of machine learning algorithms to improve the accuracy and intelligence of serial and parallel processing; model training is the continuous training and optimization of models to adapt to new data and business needs.

[0057] Data mining algorithm integration includes algorithm selection and algorithm optimization. Algorithm selection is to select a suitable data mining algorithm based on business needs; algorithm optimization is to tune the selected algorithm to improve performance and accuracy.

[0058] Key information matching includes case association and clue mining. Case association is to match personnel information with case information to identify persons involved and potential clues; clue mining is to use personnel relationships and case information to dig out deep-level crime clues.

[0059] F. Accurately push abnormal objects.

[0060] Accurate push of exception objects includes serial and parallel generation, serial and parallel push, push and feedback, visual display, real-time update and dynamic matching, and reminder mechanism.

[0061] String and parallel generation includes graph construction and gang identification. Graph construction is to build a relationship map of people and organizations to clearly show the relationship between individuals; gang identification is to identify closely connected gangs or organizations through graph analysis.

[0062] Serial and parallel push includes automatic push and customized push. Automatic push is to automatically push the serial and parallel results to relevant business systems or case handlers; customized push is to customize the push content and push timing according to user needs.

[0063] Push and feedback include group push and feedback mechanisms. Group push is to push the intelligently strung together group information to the case-handling unit; the feedback mechanism is for the case-handling unit to sign for and provide feedback on the pushed group information to verify the accuracy of the information.

[0064] Visual display includes chart display and interactive functions. Chart display uses charts to display personnel string and parallel results, such as network diagrams, tree diagrams, etc.; interactive functions provide interactive functions that allow users to explore different nodes and relationships in depth.

[0065] Real-time update and dynamic matching include real-time data stream and incremental matching. Real-time data stream processes real-time data stream and dynamically updates matching results. Incremental matching performs incremental matching on newly added data instead of re-matching the entire data set.

[0066] The reminder mechanism includes a real-time system and a notification system. The real-time reminder is to remind relevant personnel in real time when key information is updated or new clues appear; the notification system is to establish a notification system to ensure that information can be conveyed to relevant personnel in a timely manner.

[0067] Through comprehensive analysis of historical events of the XX incident and related personnel who have been arrested, we can find out a large number of available features that are closely related to the characteristics of the XX incident and related personnel and can determine potential related events; for example, by analyzing the age, ethnicity, marital status, residence status, high-risk household registration, whether the occupation is legitimate, criminal record matching, related circumstances, activities at the time of the incident, distance from the specific scene, abnormal behavior before and after the incident, and closely related persons of the personnel, we can determine the support and confidence of the matching association between the features and the event.

[0068] For the typical features that can determine whether the historically relevant personnel are involved in the case, the Apriori association rule is used to find some important frequent item sets, and the decisive key item set with the largest number of factors is found by connecting the frequent item sets. The factors in the key item set are the features of whether the selected personnel are potentially involved in the case.

[0069] After selecting the key features, the decision tree algorithm is used to simulate decisions on subsequent new event data to analyze whether specific personnel have related abnormal behaviors.

[0070] The organizational structure is guided by the actual business needs of users, takes personnel serial and parallel business as the entry point, uses big data and cloud computing as technology, and aims at intelligent and precise serial and parallel. It is formed by integrating and optimizing multiple data mining algorithms and combining them with automated processing. The processing process of this method mainly includes the steps of data reading, OCR recognition, role comparison, correlation calculation, serial and parallel generation, serial and parallel push, and reminder.

[0071] The present invention is based on new technologies such as cloud computing and big data, makes full use of the existing big data resource system, and constructs an intelligent mining and analysis model based on historical events through algorithms and means such as data mining and machine learning. It integrates the latest machine learning and deep learning algorithms to improve the accuracy of fraud behavior identification, and automatically learns historical case investigation experience, intelligently analyzes and infers, and accurately pushes relevant objects of new events to provide staff with intelligent investigation application scenarios, which solves the innovative transformation from passive investigation and analysis of traditional staff to active investigation and push of clues by new machines, and realizes a new situation of smart investigation and intelligent handling application. It has real-time updating and self-learning capabilities, and the design system can update relevant databases in real time, and has the ability to self-learn new means. It uses scientific and technological means to improve the efficiency of grassroots staff in investigation and handling, and basically realizes the beautiful vision of "staying at home and being able to solve cases all over the world", comprehensively improves the ability and level of public security behavior investigation and judgment, better assists grassroots staff in conducting accurate analysis and judgment when facing complex events, and improves the efficiency and accuracy of related event investigation.

Claims

1. A personnel feature stringing and paralleling method based on an organizational chart, characterized by: It includes the following steps: A. Select typical cases from the historical analysis and judgment factors of relevant personnel, and input historical relevant personnel factors for sample training; B. Analyze and generate key element characteristics; C. Form a decision tree of key elements and related objects; D. Enter a new set of related object conditions; E. Intelligent mining and analysis based on decision tree; F. Accurately push abnormal objects.

2. The personnel feature serial and parallel method based on the organizational chart according to claim 1 is characterized by: In step A, sample training includes data collection and information integration. Data collection includes organizational structure information and basic personnel information. Information integration includes multi-source data integration and information cleaning.

3. The personnel feature serial and parallel method based on the organizational chart according to claim 1 is characterized by: In step B, the key elements of analysis include data preprocessing, key field recognition, fuzzy matching technology and OCR recognition.

4. The personnel feature serial and parallel method based on the organizational chart according to claim 1 is characterized by: In step C, the decision tree of key elements and related objects includes intelligent concatenation and parallelization, rule engine, machine learning algorithm, relationship network construction, intersection and correlation point analysis, and search and indexing technology.

5. The personnel feature serial and parallel method based on the organizational chart according to claim 1 is characterized by: In the step D, setting the object condition set includes data reading, role comparison and organizational chart drawing.

6. The personnel feature serial and parallel method based on the organizational chart according to claim 1 is characterized by: In step E, the decision tree intelligent mining analysis includes correlation calculation, machine learning algorithm, big data technology, cloud computing platform, intelligent and precise serial and parallel, data mining algorithm integration and key information matching.

7. The personnel feature serial and parallel method based on the organizational chart according to claim 1 is characterized by: In the step F, accurate push of abnormal objects includes serial-parallel generation, serial-parallel push, push and feedback, visual display, real-time update and dynamic matching, and reminder mechanism.

Images