Worker compliance assessment system based on big data mining

Abstract

This invention discloses an employment compliance assessment system based on big data mining, relating to the field of big data technology. It preprocesses scattered contract information, business management information, process traceability information, and entity association information to form an employment event set Evt. It further extracts the contract entity set Con, the performance entity set Ful, and the registration entity set Reg. Then, it combines these with the subsequently formed control relationship chain result Ctl, the actual management entity result Rst, and the entity deviation result Dev for judgment. This system can organize the originally scattered and fragmented contract relationships, performance relationships, and registration relationships into a single traceable and verifiable management relationship chain. Thus, without increasing hardware requirements, it can identify the risk of inconsistency between the nominal entity and the actual management entity. For example, if a business is registered under an outsourcing unit, but scheduling, training distribution, and penalty processing are continuously initiated by store managers, problems can be detected before disputes arise.

Description

Technical Field

[0001] This invention relates to the field of big data technology, specifically to an employment compliance assessment system based on big data mining. Background Technology

[0002] As employment scenarios such as chain retail, platform delivery, property services, warehousing and distribution operations, and manufacturing collaboration continue to expand, enterprises have generated a large amount of business data, log data, and process data in the processes of personnel organization, task allocation, shift arrangement, training distribution, performance appraisal, reward and punishment implementation, and remuneration confirmation. The above-mentioned multi-source data provides a continuously accumulating data foundation for employment risk identification and compliance status assessment.

[0003] Currently, with enterprises employing a mix of formal employees, dispatched workers, outsourced on-site staff, hourly workers, and temporary workers, although they have deployed electronic contract systems, scheduling systems, approval systems, training systems, payroll systems, and social insurance reporting systems, most traditional solutions still rely on fragmented judgments based on contract texts, attendance results, wage results, or turnover probabilities. They struggle to delve deeper into scheduling release records, task assignment records, approval records, training distribution records, penalty records, and wage confirmation records to uncover the true management control relationships: who issued the actual management actions, who controlled the actual work arrangements, who executed the actual assessments and punishments, and who confirmed the actual compensation. This makes it difficult to accurately identify discrepancies between the nominal contracting entity, the wage disbursement entity, the social insurance payment entity, and the actual management entity. Consequently, it fails to promptly detect hidden risks such as nominal outsourcing but actual dispatch, or nominal dispatch but actual direct employment. Existing technologies still have significant shortcomings in penetrating and identifying employment responsibilities. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an employment compliance assessment system based on big data mining, which solves the problems mentioned in the background section.

[0005] This invention is achieved through the following technical solution: an employment compliance assessment system based on big data mining, including a data collection module, an action recognition module, a relationship restoration module, a subject verification module, and an assessment output module; The data collection module is used to collect contract information, business management information, process trace information and subject association information associated with the target employment object, and preprocess them to form an employment event set Evt, including a contract subject set Con, a performance subject set Ful and a registration subject set Reg. The action recognition module is used to perform management action recognition on the event records in the employment event set Evt, extract the action initiator, action receiver, action type and action occurrence time, and form a management action set Act; The relationship restoration module is used to construct the control relationship chain result Ctl based on the temporal correlation and business connection relationship between each management action record in the management action set Act, and to identify the actual management entity result Rst corresponding to the target employment object based on the continuous control status of each entity in multiple management links in the control relationship chain result Ctl. The subject verification module is used to cross-verify the actual management subject result Rst with the contract subject set Con, the performance subject set Ful, and the registration subject set Reg, respectively, to form a subject deviation result Dev, which represents the deviation between the actual management subject result Rst and the contract subject set Con, the performance subject set Ful, and the registration subject set Reg. The evaluation output module is used to determine the employment relationship and assess the risk based on the control relationship chain result Ctl and the subject deviation result Dev, and output the evaluation result.

[0006] Preferably, the data collection module includes an information compilation unit; The information compilation unit collects contract information, business management information, process trace information, and subject association information associated with the target employment object, and performs cleaning, association mapping, and time sorting processing on the contract information, business management information, process trace information, and subject association information to form a collection record arranged continuously in the order of occurrence. After integration, a collection record sequence Rec is formed.

[0007] Preferably, the data collection module further includes an event modeling and extraction unit; The event modeling and extraction unit performs event transformation processing based on the aggregated records in the aggregated record sequence Rec to form an employment event set Evt. And extract the subject information representing the contractual relationship from the employment event set Evt to form the contract subject set Con; Extract the subject information representing the actual performance relationship to form a set of performing subjects, Ful; Extract the main information representing the registration relationship to form a set of registration entities, Reg.

[0008] Preferably, the action recognition module includes an event screening unit and an action generation unit; The event screening unit reads the event records in the employment event set Evt one by one and determines whether the current event record represents a management and control behavior initiated by a certain subject on the target employment object; When the current event record contains any of the following event contents: task arrangement, shift arrangement, approval control, training distribution, assessment processing, penalty processing, remuneration confirmation, job adjustment, or access control, the action initiator, action receiver, action type, and action occurrence time are extracted from the current event record to generate the corresponding candidate action record Acr; then the generated candidate action record Acr is written into the action candidate set Acu.

[0009] Preferably, the action recognition module further includes an action generation unit; The action generation unit reads each candidate action record Acr from the action candidate set Acu, and merges and organizes the action initiator, action receiver, action type, and action occurrence time in each candidate action record Acr; when multiple candidate action records Acr correspond to the same action initiator, the same action receiver, and the same action type, and the action occurrence time is consistent or continuously associated, the above candidate action records Acr are merged. Then, the organized action records are written into the management action set Act in chronological order of the actions occurring.

[0010] Preferably, the relationship restoration module includes a link construction unit; The link construction unit reads each management action record Atr in the management action set Act one by one, and then arranges them from earliest to latest according to the time of occurrence of the action corresponding to each management action record Atr, so as to obtain the action order corresponding to the same target user. Then, compare the two adjacent management action records (ATr) one by one to determine whether there is a temporal succession relationship and a business succession relationship between the two management action records (ATr). Among them, if the receiving entity of the current management action record Atr is the same as the initiating entity of the next management action record Atr, it is determined that there is a subject succession relationship between the two management action records Atr. When the action types of two consecutive management action records (ATr) belong to consecutive links in the same management process, it is determined that there is a business connection relationship between the two consecutive management action records (ATr). When either the main body undertaking relationship or the business connection relationship is met, the two management action records (Atr) are connected as adjacent nodes in the same control chain. The join process continues until the control relationship chain result Ctl, arranged sequentially in chronological order, is obtained.

[0011] Preferably, the relationship restoration module further includes a subject identification unit; The subject identification unit reads the link content in the control relationship chain result Ctl one by one, and counts the number of times each subject appears in the same control relationship chain result Ctl, the consecutive occurrence segments, and the distribution of the corresponding management links; When the same entity continuously initiates management actions in multiple management stages, or repeatedly appears as the initiator of management actions in multiple adjacent link nodes, the entity is identified as the entity that continuously controls the target employment object; then, the identified entity is written into the actual management entity result Rst.

[0012] Preferably, the subject verification module includes a subject comparison unit; The subject comparison unit reads the actual management subject result Rst, the contract subject set Con, the performance subject set Ful, and the registration subject set Reg corresponding to the target employment object; The actual management entity in the actual management entity result Rst is compared item by item with the contract entity in the contract entity set Con, the performance entity in the performance entity set Ful, and the registration entity in the registration entity set Reg; When the actual management entity in the actual management entity result Rst is the same as or mapped to the same entity in the corresponding set, the result is recorded as consistent; When the actual management entity in the actual management entity result Rst is different from the entity in the corresponding set, and cannot be mapped to the same entity, the result is recorded as inconsistent; Then, write the comparison results of each item into the corresponding subject verification record Vfy.

[0013] Preferably, the subject verification module further includes a deviation determination unit; The deviation determination unit reads each subject verification record Vfy and makes a deviation determination based on the comparison results recorded in each subject verification record Vfy. When the actual management entity result Rst is consistent with the contract entity set Con, the performance entity set Ful, and the registration entity set Reg, it is determined to be without deviation; When the actual management entity's result Rst is consistent with only a portion of the set, it is judged as a partial deviation; When the actual management entity result Rst is inconsistent with the contract entity set Con, the performance entity set Ful, and the registration entity set Reg, it is judged as a significant deviation; Then write the judgment result into the main deviation result Dev.

[0014] Preferably, the evaluation output module includes a comprehensive evaluation unit; The comprehensive assessment unit reads the control relationship chain result Ctl and the subject deviation result Dev corresponding to the target employee, and performs employment relationship determination and compliance risk assessment based on the number of times the same subject appears in the control relationship chain result Ctl in task arrangement, shift arrangement, approval control, training distribution, assessment processing, penalty processing, remuneration confirmation, job adjustment and access control, and the deviation judgment result recorded in the subject deviation result Dev. When the same entity appears as the action initiator in at least three of the above management stages in the control relationship chain result Ctl, and the entity deviation result Dev is a significant deviation, the corresponding target employment object is determined to be a subject misalignment relationship, and the corresponding relationship determination result Jud is output. At the same time, the corresponding target employment object is assessed as a high-risk state, and the corresponding compliance assessment result Eva is output. When the same entity appears as the action initiator in at least two of the above management stages in the control relationship chain result Ctl, and the entity deviation result Dev is a partial deviation, the corresponding target user is determined to be in a subject separation relationship, and the corresponding relationship determination result Jud is output. At the same time, the corresponding target user is assessed as a medium-risk state, and the corresponding compliance assessment result Eva is output. When the same subject appears as the action initiator only in one management link in the control relationship chain result Ctl, and the subject deviation result Dev is no deviation, the corresponding target user is determined to be a subject consistency relationship, and the corresponding relationship determination result Jud is output. At the same time, the corresponding target user is assessed as a low-risk state, and the corresponding compliance assessment result Eva is output.

[0015] This invention provides an employment compliance assessment system based on big data mining, which has the following beneficial effects: (1) By preprocessing the scattered contract information, business management information, process trace information, and subject association information, an employment event set Evt is formed. The contract subject set Con, the performance subject set Ful, and the registration subject set Reg are further extracted. Then, the control relationship chain result Ctl, the actual management subject result Rst, and the subject deviation result Dev are combined to make a judgment. This can organize the originally scattered and fragmented contract relationships, performance relationships, and registration relationships into a single traceable and verifiable management relationship chain, thereby identifying the risk of inconsistency between the nominal subject and the actual management subject without increasing hardware. For example, when a store salesperson is registered under the name of an outsourced unit, but the scheduling, training distribution, and penalty handling are continuously initiated by the store management personnel, the problem of inconsistency between the registration relationship and the actual management relationship can be discovered before a dispute occurs.

[0016] (2) By filtering out event content that can represent management control behavior from the employment event set Evt, generating candidate action records Acr, and then merging them to form a management action set Act, the duplicate records of the same management behavior in different systems can be compressed into a single real management action, avoiding misjudging a shift adjustment, an approval control, or a training distribution as multiple independent management behaviors. For example, when a shift adjustment generates release records, confirmation records, and process trace records at the same time, the management action set Act formed after merging is closer to the actual number and sequence of management controls, thereby reducing the interference of duplicate records on subsequent judgments.

[0017] (3) By connecting the management action records Atr in the management action set Act according to the order of business occurrence and management succession relationship into a control relationship chain result Ctl, and further identifying the actual management subject result Rst, the originally discrete management actions can be restored into a continuous management control process, and the subject that truly and continuously issues control behavior can be identified. For example, when a certain on-site personnel first accepts the shift arrangement, and then accepts the approval control, training distribution and assessment processing, multiple discrete actions can be connected into the same control relationship chain result Ctl, and the subject that continuously dominates the above management links can be identified, thus providing a more realistic judgment basis for the subsequent formation of the subject deviation result Dev and the output compliance assessment result Eva. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the overall structure of the employment compliance assessment system based on big data mining according to the present invention; Figure 2 A schematic diagram illustrating data collection and event modeling; Figure 3 This is a schematic diagram of action recognition and merging processing; Figure 4 This is a schematic diagram for relationship reconstruction and subject identification. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0020] Example 1 This invention provides an employment compliance assessment system based on big data mining. Please refer to [link / reference]. Figures 1 to 4 It includes a data collection module, an action recognition module, a relationship restoration module, a subject verification module, and an evaluation output module; The data collection module is used to collect contract information, business management information, process trace information and subject association information associated with the target employment object, and preprocess them to form an employment event set Evt, including a contract subject set Con, a performance subject set Ful and a registration subject set Reg. The action recognition module is used to perform management action recognition on the event records in the employment event set Evt, extract the action initiator, action receiver, action type and action occurrence time, and form a management action set Act; The relationship restoration module is used to construct the control relationship chain result Ctl based on the temporal correlation and business connection relationship between each management action record in the management action set Act, and to identify the actual management entity result Rst corresponding to the target employment object based on the continuous control status of each entity in multiple management links in the control relationship chain result Ctl. The subject verification module is used to cross-verify the actual management subject result Rst with the contract subject set Con, the performance subject set Ful, and the registration subject set Reg, respectively, to form a subject deviation result Dev, which represents the deviation between the actual management subject result Rst and the contract subject set Con, the performance subject set Ful, and the registration subject set Reg. The evaluation output module is used to determine the employment relationship and assess the risk based on the control relationship chain result Ctl and the subject deviation result Dev, and output the evaluation result.

[0021] In this embodiment, the data collection module first preprocesses the contract information, business management information, process trace information, and subject association information that were originally scattered in different systems to form an employment event set Evt. At the same time, it extracts the contract subject set Con, the performance subject set Ful, and the registration subject set Reg. Then, the action recognition module further organizes the scattered events into a management action set Act that can directly represent management behavior. Next, the relationship restoration module constructs a control relationship chain result Ctl based on the management action set Act and identifies the actual management subject result Rst. Finally, the subject verification module generates the subject deviation result Dev, and the evaluation output module outputs the evaluation result. Through the above processing, the scattered records in the prior art, which can only see the contract relationship, performance relationship, and registration relationship separately but cannot see who actually manages the target employment object, can be transformed into a management control link that can be continuously traced, cross-verified, and directly determined. This solves the problem mentioned in the background that it is difficult to identify the nominal contracting subject, nominal performance subject, or nominal registration subject in a timely manner when they are inconsistent with the actual management subject. Specifically, when chain retail enterprises register their store employees under outsourced units, but the scheduling, training, approval, penalties, and job adjustments are all initiated continuously by the store's direct management personnel, the control relationship chain result Ctl can connect these multiple management links. The actual management entity result Rst can identify the entity that truly and continuously issues control actions. Combined with the contract entity set Con, the performance entity set Ful, and the registration entity set Reg, the entity deviation result Dev is formed, thus identifying the risk of inconsistency between the registered relationship and the actual management relationship before disputes arise. Therefore, the special benefit of the above solution is not just completing general information collection, but also restoring the actual management relationship hidden in multi-source business records without increasing hardware. It also concretely translates the misalignment between formal and actual relationships into verifiable, explainable, and risk-assessable results, thereby reducing problems such as unclear employment responsibility, long-term risk accumulation, and passive post-event evidence collection.

[0022] Example 2 Please see Figure 2 Specifically: the data collection module includes an information compilation unit; The information compilation unit collects contract information, business management information, process trace information, and subject association information associated with the target employment object, and performs cleaning, association mapping, and time sorting processing on the contract information, business management information, process trace information, and subject association information to form a collection record arranged continuously in the order of occurrence. After integration, a collection record sequence Rec is formed. It should be noted that: The contract information refers to the contract records associated with the target employer in the contract management system or file storage location. The contract information includes at least one or more of the following: the name of the contracting party, the identifier of the target employer, the contract formation time, the contract effective time, the contract expiration time, the contract status, and the contract source location. The business management information refers to management records generated by the target employees during the processes of scheduling, work assignment, approval, training, assessment, rewards and punishments, job adjustment and remuneration confirmation. The business management information includes at least one or more of the following: the action initiator, the action recipient, the action type, the time of action occurrence and the location of the business source. The process trace information is an operation trace record formed by the target employee during the business process. The process trace information includes at least one or more of the following: operation subject, operation time, status change content, confirmation content, and trace source location. The subject association information is a record of the attribution, connection, registration or mapping relationship between the target employment object and different subjects. The subject association information includes at least one or more of the following: subject name, subject category, target employment object identifier, association relationship type, association effective time and association expiration time. Performing cleaning, association mapping, and time sorting processes means: first, reading the contract information, business management information, process trace information, and subject association information one by one, and deleting invalid records that lack target employment object identifiers, lack subject names and cannot be completed, or lack time information and cannot be completed; Next, the time fields in records from different sources are unified to the same time format, the subject names in records from different sources are unified to the same subject name, and the action names in records from different sources are unified to the same action type, thus completing the cleaning process. After that, records corresponding to the same target user are merged into the same processing chain based on the target user identifier, and records representing the same user in different systems are mapped to the same user item based on the subject name, subject category, and association relationship type, thus completing the association mapping. Finally, the records are sorted first by the time of the business occurrence. When the time of the business occurrence is missing, the records are sorted by the business effective time, status change time, or record generation time in turn. When two records have the same sorting time, they are sorted in the order of contract records, management records, trace records, and related records, thus completing the time sorting process and obtaining a collection record sequence Rec arranged continuously in the order of occurrence.

[0023] The data collection module also includes an event modeling and extraction unit; The event modeling and extraction unit performs event transformation processing based on the aggregated records in the aggregated record sequence Rec to form an employment event set Evt. And extract the subject information representing the contractual relationship from the employment event set Evt to form the contract subject set Con; Extract the subject information representing the actual performance relationship to form a set of performing subjects, Ful; Extract the entity information representing the registration relationship to form a registration entity set Reg; It should be noted that: Based on the event transformation processing of the aggregated records in the aggregated record sequence Rec, the employment event set Evt is formed by: reading the aggregated records in the aggregated record sequence Rec one by one, and first determining which category the current aggregated record belongs to among contract formation, contract modification, contract termination, management arrangement, management execution, status recording, registration establishment, registration modification, or registration cancellation; Then convert the current collection record into a corresponding event record, and write the target employment object identifier, event type, event occurrence time, event associated subject, event source location and event content summary into the event record; The event occurrence time is preferentially adopted from the business occurrence time in the collection records. If the business occurrence time is missing, the business effective time, status change time, or record generation time are adopted in sequence. After the conversion of all collection records is completed, the event records are arranged in order of event occurrence time to obtain the employment event set Evt.

[0024] Extracting the subject information representing the contractual relationship from the employment event set Evt to form the contract subject set Con means: selecting event records from the employment event set Evt whose event type is contract formation, contract modification, and contract termination; Then read the name of the contracting entity or the name of the entity corresponding to the contract from the above event records; when multiple event records correspond to the same entity, only one entity record is retained; when the name of the same entity is written differently in different event records but corresponds to the same entity identifier, they are merged into the same entity record; after deduplication and merging, a set of contract entities Con is formed.

[0025] Extracting the subject information representing the actual performance relationship from the employment event set Evt to form the performance subject set Ful means: filtering event records from the employment event set Evt that are of event type management arrangement, management execution, and status tracking; then reading the action initiator name, management corresponding subject name, or execution control subject name from the above event records; when the same subject appears repeatedly in multiple event records, only one subject record is retained; when the same subject appears consecutively in multiple event records, they are still merged into the same subject record; after deduplication and merging, the performance subject set Ful is formed.

[0026] Extracting the entity information representing the registration relationship from the employment event set Evt to form the registration entity set Reg means: filtering event records with event types of registration establishment, registration modification, and registration cancellation from the employment event set Evt; then reading the registration corresponding entity name, the attached corresponding entity name, or the belonging corresponding entity name from the above event records; when multiple event records correspond to the same entity, only one entity record is retained; after deduplication and merging, the registration entity set Reg is formed.

[0027] In this embodiment, through the above processing, the data collection module first uses the information compilation unit to pull the contract information, business management information, process trace information, and subject association information, which were originally scattered in the contract management system, scheduling system, approval system, training system, reward and punishment system, and registration linkage system, into the same processing link. Then, by deleting invalid records, unifying the time format, unifying the subject name, unifying the action type, and merging and sorting according to the same target employment object, a collection record sequence Rec is formed in the order of occurrence. On this basis, the event modeling extraction unit then converts the collection records in the collection record sequence Rec into a set of employment events Evt with consistent format and continuous time, and further extracts them to form a contract subject set Con, a performance subject set Ful, and a registration subject set Reg. The unique advantage of this is that this process does not directly determine who is actually managing the target employees. Instead, it first organizes the underlying business records, which were originally inconsistent in writing, source, time, and content, into a standardized event foundation that can be continuously read, traced back and verified, and directly used for subsequent identification and processing. This solves the problem of fragmented records from different systems, multiple names for the same entity, and chaotic order of the same business action in existing technologies, which leads to subsequent identification distortion. In a real-world scenario, for example, the same salesperson in a chain store might be listed as the full name of the outsourcing company in the contract record, the abbreviation of the store in the scheduling record, the project team name in the training record, and the supplier code in the registration record. Without first forming a sequence of aggregated records (Rec) through an information processing unit, it would be easy to mistakenly identify the same entity as multiple entities. However, after the above processing, not only can we obtain a set of employment events (Evt) with a clear time sequence, but we can also simultaneously obtain a set of contract entities (Con), a set of performing entities (Ful), and a set of registration entities (Reg) representing contractual relationships, actual performance relationships, and registration relationships, respectively. This provides a consistent input basis for subsequent modules, ensuring consistency in source, time, and entity, and preventing subsequent judgments from being based on chaotic data.

[0028] Example 3 Please see Figure 3 Specifically: the action recognition module includes an event screening unit and an action generation unit; The event screening unit reads the event records in the employment event set Evt one by one and determines whether the current event record represents a management and control behavior initiated by a certain subject on the target employment object; When the current event record contains any of the following event contents: task arrangement, shift arrangement, approval control, training distribution, assessment processing, penalty processing, remuneration confirmation, job adjustment, or access control, the action initiator, action receiver, action type, and action occurrence time are extracted from the current event record to generate the corresponding candidate action record Acr; then the generated candidate action record Acr is written into the action candidate set Acu. Each candidate action record (Acr) includes at least the action initiator, action receiver, action type, action occurrence time, and action source information; It should be noted that: Extracting the initiator, receiver, type, and time of action refers to determining the initiator, receiver, type, and time of occurrence of a management action from the current event log. The action type is determined by the event category to which the current event record belongs; The time of an action is determined by the business occurrence time in priority. If the business occurrence time is missing, the business effective time, status change time, or record generation time are used in turn to determine the time. Once the above information is determined, the corresponding candidate action record Acr is generated.

[0029] The action recognition module also includes an action generation unit; The action generation unit reads each candidate action record Acr from the action candidate set Acu, and merges and organizes the action initiator, action receiver, action type, and action occurrence time in each candidate action record Acr; when multiple candidate action records Acr correspond to the same action initiator, the same action receiver, and the same action type, and the action occurrence time is consistent or continuously associated, the above candidate action records Acr are merged. The organized action records are then written into the management action set Act in chronological order of the actions, forming the management action set Act for subsequent relationship restoration processing; It should be noted that: The merging process refers to: reading each candidate action record Acr in the action candidate set Acu; when two or more candidate action records Acr correspond to the same action initiator, the same action receiver, the same action type, and the same action occurrence time or the interval between them does not exceed the preset merging time, merging the two or more candidate action records Acr into one action record. During the merging process, the earliest action occurrence time is retained as the action occurrence time after the merge, and the corresponding source information is written into the merged action record. When two or more candidate action records Acr do not simultaneously meet the above conditions, they remain as independent action records and are not merged.

[0030] Meanwhile, the preset merging time is a continuous judgment time preset according to the same business type; in implementation, corresponding preset merging times can be set for task arrangement, shift arrangement, approval control, training distribution, assessment processing, penalty processing, remuneration confirmation, job adjustment and access control respectively.

[0031] In this embodiment, the special benefit of the action recognition module lies not in continuing to supplement the underlying data, but in further filtering out the contract events, record events, registration events and management events that were originally mixed in the employment event set Evt, and extracting the event content that can truly indicate who initiated management control. The above event content is then uniformly compressed into candidate action records Acr and management action sets Act with consistent format, clear time and traceable source. This solves the problem in the prior art that there are many management behavior records, but very few action records that can be used to determine control relationships, and the same management behavior is repeatedly recorded due to system push, status write-back, confirmation record, etc. In real-world scenarios, such as a shift adjustment in a chain retail store, multiple records are often generated simultaneously: shift release records, message push records, employee confirmation records, and process log records. Without processing by the event screening unit and the action generation unit, these multiple records can easily be mistaken for multiple independent control actions. This can lead to the amplification of a single real management action into multiple management actions when constructing the control chain. However, after the event screening unit filters out management control events and generates candidate action records (Acr), the action generation unit performs a merging process based on the action initiator, receiver, type, and time of occurrence. The content ultimately written into the management action set (Act) more closely approximates the actual number and sequence of management controls. Therefore, when subsequent modules read the management action set (Act), they not only reduce interference from repetitive actions in determining control relationships but also retain the earliest action occurrence time and corresponding source information. This makes the starting point of management actions clearer, the boundaries more stable, and the time sequence more accurate. This benefit differs from the benefit of organizing scattered underlying records into a unified event base. The action recognition module further addresses the problem of extracting real management actions from a unified event base and removing repetitive interference.

[0032] Example 4 Please see Figure 4 Specifically: the relationship restoration module includes a link construction unit; The link construction unit reads each management action record Atr in the management action set Act one by one, and then arranges them from earliest to latest according to the time of occurrence of the action corresponding to each management action record Atr, so as to obtain the action order corresponding to the same target user. Then, compare the two adjacent management action records (ATr) one by one to determine whether there is a temporal succession relationship and a business succession relationship between the two management action records (ATr). Among them, if the receiving entity of the current management action record Atr is the same as the initiating entity of the next management action record Atr, it is determined that there is a subject succession relationship between the two management action records Atr. When the action types of two consecutive management action records (ATr) belong to consecutive links in the same management process, it is determined that there is a business connection relationship between the two consecutive management action records (ATr). When either the main body undertaking relationship or the business connection relationship is met, the two management action records (Atr) are connected as adjacent nodes in the same control chain. The connection processing continues until the control relationship chain result Ctl, which is arranged continuously in chronological order, is obtained. Through the above processing, the originally scattered management action record Atr can be restored into a complete management control transmission link. It should be noted that: Continuing to perform connection processing means that after connecting two management action records Atr that meet the connection conditions into the same control chain, the next management action record Atr that is in the current chain tail record is used as the next chain tail record. Then, the next management action record Atr that is in the current chain tail record in terms of time sequence is read, and it is determined whether the next management action record Atr read satisfies the main body inheritance relationship or business connection relationship with the current chain tail record. When the conditions are met, the next management action record (ATr) read is appended to the end of the current control chain, and the newly appended management action record (ATr) is updated to the current chain tail record. Then, continue reading the next management action record (Atr) that is in the time sequence after the updated current chain tail record, repeat the above judgment and append processing, until all subsequent management action records (Atr) no longer meet the connection conditions.

[0033] Meanwhile, when the subsequent management action record Atr does not meet the connection conditions, the extension of the current control chain is stopped, and the entire chain that has been connected is written into the control relationship chain result Ctl; Then, from the remaining management action records Atr that have not yet been written to the control relationship chain result Ctl, select the earliest management action record Atr as the new starting record, and perform the connection processing in the same way.

[0034] The relationship restoration module also includes a subject identification unit; The subject identification unit reads the link content in the control relationship chain result Ctl one by one, and counts the number of times each subject appears in the same control relationship chain result Ctl, the consecutive occurrence segments, and the distribution of the corresponding management links; When the same entity continuously initiates management actions in multiple management stages, or repeatedly appears as the initiator of management actions in multiple adjacent link nodes, the entity is identified as the entity that continuously controls the target employment object; then, the identified entity is written into the actual management entity result Rst. Through the above processing, the entity that continuously issues control actions in the actual business process can be identified from the control relationship chain result Ctl, and it can be used as the actual management entity output for subsequent cross-verification with the contract entity set Con, the performance entity set Ful, and the registration entity set Reg.

[0035] In this embodiment, the unique advantage of the relationship restoration module goes beyond the data organization or deduplication of management actions described earlier. It further restores the existing set of management actions (Act) into a continuously traceable control relationship chain result (Ctl) according to the actual business occurrence sequence and management acceptance logic. Based on this, the actual management entity result (Rst) is output. This solves the problem in existing technologies where, although individual management action records can be seen, it is still impossible to determine whether these records are continuously dominated by the same entity, or whether the control behavior is sporadic or has formed a stable management chain. Specifically, in real-world employment scenarios, a site employee might first have their shift arranged by the store duty manager, then approved by the regional supervisor, then have training requirements issued by the store training manager, and finally have performance evaluations or penalties implemented by store management personnel. If only individual management action records (Atr) are viewed one by one, only multiple discrete management behaviors can be seen, making it difficult to confirm whether these behaviors belong to the same continuous control chain, and also difficult to identify who has formed continuous dominant control over the target employee. The link construction unit connects management action records (Atr) with corresponding subject-subject relationships or business connections into a single control relationship chain result (Ctl). The subject identification unit then statistically analyzes the frequency of the same subject's appearance in multiple management stages, consecutive occurrence segments, and the distribution of management stages. This allows the scattered management actions to be reconstructed into a complete management control chain, further identifying the subject that consistently issues control actions and forming the actual management subject result (Rst). The direct benefit of this is that the system no longer merely observes what actions occur, but can further clarify how these actions are continuously transmitted and who ultimately maintains control. This enables subsequent cross-verification of the contract subject set (Con), the performance subject set (Ful), and the registration subject set (Reg) to be based on the real management chain. This is particularly suitable for uncovering hidden problems in chain stores, platform employment, or on-site service scenarios where the apparent subject and the actual controlling subject are separated for a long period, but individual records are not readily apparent.

[0036] Example 5 Specifically: the subject verification module includes a subject comparison unit; The subject comparison unit reads the actual management subject result Rst, the contract subject set Con, the performance subject set Ful, and the registration subject set Reg corresponding to the target employment object; The actual management entity in the actual management entity result Rst is compared item by item with the contract entity in the contract entity set Con, the performance entity in the performance entity set Ful, and the registration entity in the registration entity set Reg; When the actual management entity in the actual management entity result Rst is the same as or mapped to the same entity in the corresponding set, the result is recorded as consistent; When the actual management entity in the actual management entity result Rst is different from the entity in the corresponding set, and cannot be mapped to the same entity, the result is recorded as inconsistent; Then, write the comparison results of each item into the corresponding subject verification record Vfy.

[0037] The main verification module also includes a deviation determination unit; The deviation determination unit reads each subject verification record Vfy and makes a deviation determination based on the comparison results recorded in each subject verification record Vfy. When the actual management entity result Rst is consistent with the contract entity set Con, the performance entity set Ful, and the registration entity set Reg, it is determined to be without deviation; When the actual management entity's result Rst is consistent with only a portion of the set, it is judged as a partial deviation; When the actual management entity result Rst is inconsistent with the contract entity set Con, the performance entity set Ful, and the registration entity set Reg, it is judged as a significant deviation; Then write the judgment result into the main deviation result Dev.

[0038] The evaluation output module includes a comprehensive evaluation unit; The comprehensive assessment unit reads the control relationship chain result Ctl and the subject deviation result Dev corresponding to the target employee, and performs employment relationship determination and compliance risk assessment based on the number of times the same subject appears in the control relationship chain result Ctl in task arrangement, shift arrangement, approval control, training distribution, assessment processing, penalty processing, remuneration confirmation, job adjustment and access control, and the deviation judgment result recorded in the subject deviation result Dev. When the same entity appears as the action initiator in at least three of the above management stages in the control relationship chain result Ctl, and the entity deviation result Dev is a significant deviation, the corresponding target employment object is determined to be a subject misalignment relationship, and the corresponding relationship determination result Jud is output. At the same time, the corresponding target employment object is assessed as a high-risk state, and the corresponding compliance assessment result Eva is output. When the same entity appears as the action initiator in at least two of the above management stages in the control relationship chain result Ctl, and the entity deviation result Dev is a partial deviation, the corresponding target user is determined to be in a subject separation relationship, and the corresponding relationship determination result Jud is output. At the same time, the corresponding target user is assessed as a medium-risk state, and the corresponding compliance assessment result Eva is output. When the same subject appears as the action initiator only in one management link in the control relationship chain result Ctl, and the subject deviation result Dev is no deviation, the corresponding target user is determined to be a subject consistency relationship, and the corresponding relationship determination result Jud is output. At the same time, the corresponding target user is assessed as a low-risk state, and the corresponding compliance assessment result Eva is output.

[0039] In this embodiment, the special advantage of the subject verification module and the assessment output module through the above processing is that, based on the actual management subject result Rst and the control relationship chain result Ctl, it no longer stops at identifying who is continuously managing, but further compares the continuously managing subject with the contract subject set Con, the performance subject set Ful, and the registration subject set Reg item by item, and solidifies the comparison result into the subject deviation result Dev. Then, combined with the number of management links actually covered by the same subject in the control relationship chain result Ctl, it directly outputs the relationship judgment result Jud and the compliance assessment result Eva. Thus, the work that originally could only rely on manual judgment of each contract, each performance record, and each registration record can be transformed into relationship judgment results and risk results that can be automatically output in layers. Specifically, in a real-world scenario, a sales associate in a shopping mall corresponds to an outsourcing company in the contract subject set Con and to a supplier in the registration subject set Reg. However, in the control relationship chain result Ctl, the mall counter supervisor appears as the action initiator in four management stages: task arrangement, shift arrangement, approval control, and penalty handling. In this case, after verifying the actual management subject result Rst against the contract subject set Con, the performance subject set Ful, and the registration subject set Reg, a significant deviation result in the subject deviation result Dev is formed. Then, the comprehensive evaluation unit directly determines the target employee as having a subject misalignment. The relationship corresponds to the output relationship judgment result Jud and the compliance assessment result Eva for high-risk status. Compared with the previous modules, which mainly solve the problems of data compilation, action extraction and link restoration, the special benefits of this part are more specifically reflected in the fact that it can put the four types of relationships—who is actually in charge, who is written on the contract, who is performed, and who is registered—into the same judgment link for unified verification, and directly establish clear relationship types and risk levels according to the established conditions. This allows enterprises to identify the specific objects and specific risks corresponding to the main body misalignment relationship, main body separation relationship, or main body consistency relationship in advance before labor disputes, inspection spot checks, or internal audits occur.

[0040] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A labor compliance assessment system based on big data mining, characterized in that: It includes a data collection module, an action recognition module, a relationship restoration module, a subject verification module, and an evaluation output module; The data collection module is used to collect contract information, business management information, process trace information and subject association information associated with the target employment object, and preprocess them to form an employment event set Evt, including a contract subject set Con, a performance subject set Ful and a registration subject set Reg. The action recognition module is used to perform management action recognition on the event records in the employment event set Evt, extract the action initiator, action receiver, action type and action occurrence time, and form a management action set Act; The relationship restoration module is used to construct the control relationship chain result Ctl based on the temporal correlation and business connection relationship between each management action record in the management action set Act, and to identify the actual management entity result Rst corresponding to the target employment object based on the continuous control status of each entity in multiple management links in the control relationship chain result Ctl. The subject verification module is used to cross-verify the actual management subject result Rst with the contract subject set Con, the performance subject set Ful, and the registration subject set Reg, respectively, to form a subject deviation result Dev, which represents the deviation between the actual management subject result Rst and the contract subject set Con, the performance subject set Ful, and the registration subject set Reg. The evaluation output module is used to determine the employment relationship and assess the risk based on the control relationship chain result Ctl and the subject deviation result Dev, and output the evaluation result.

2. The big data mining based workforce compliance evaluation system of claim 1, wherein: The data collection module includes an information compilation unit; The information compilation unit collects contract information, business management information, process trace information, and subject association information associated with the target employment object, and performs cleaning, association mapping, and time sorting processing on the contract information, business management information, process trace information, and subject association information to form a collection record arranged continuously in the order of occurrence. After integration, a collection record sequence Rec is formed.

3. The big data mining based workforce compliance evaluation system of claim 2, wherein: The data collection module also includes an event modeling and extraction unit; The event modeling and extraction unit performs event transformation processing based on the aggregated records in the aggregated record sequence Rec to form an employment event set Evt. And extract the subject information representing the contractual relationship from the employment event set Evt to form the contract subject set Con; Extract the subject information representing the actual performance relationship to form a set of performing subjects, Ful; Extract the main information representing the registration relationship to form a set of registration entities, Reg.

4. The employment compliance assessment system based on big data mining according to claim 3, characterized in that: The action recognition module includes an event screening unit and an action generation unit; The event screening unit reads the event records in the employment event set Evt one by one and determines whether the current event record represents a management and control behavior initiated by a certain subject on the target employment object; When the current event record contains any of the following event contents: task arrangement, shift arrangement, approval control, training distribution, assessment processing, penalty processing, remuneration confirmation, job adjustment, or access control, the action initiator, action receiver, action type, and action occurrence time are extracted from the current event record to generate the corresponding candidate action record Acr; then the generated candidate action record Acr is written into the action candidate set Acu.

5. The employment compliance assessment system based on big data mining according to claim 4, characterized in that: The action recognition module also includes an action generation unit; The action generation unit reads each candidate action record Acr from the action candidate set Acu, and merges and organizes the action initiator, action receiver, action type, and action occurrence time in each candidate action record Acr; when multiple candidate action records Acr correspond to the same action initiator, the same action receiver, and the same action type, and the action occurrence time is consistent or continuously associated, the above candidate action records Acr are merged. Then, the organized action records are written into the management action set Act in chronological order of the actions occurring.

6. The employment compliance assessment system based on big data mining according to claim 5, characterized in that: The relationship restoration module includes a link construction unit; The link construction unit reads each management action record Atr in the management action set Act one by one, and then arranges them from earliest to latest according to the time of occurrence of the action corresponding to each management action record Atr, so as to obtain the action order corresponding to the same target user. Then, compare the two adjacent management action records (ATr) one by one to determine whether there is a temporal succession relationship and a business succession relationship between the two management action records (ATr). Among them, if the subject receiving the action of the previous management action record Atr is the same as the subject initiating the action of the next management action record Atr, it is determined that there is a subject succession relationship between the two management action records Atr. When the action types of two consecutive management action records (ATr) belong to consecutive links in the same management process, it is determined that there is a business connection relationship between the two consecutive management action records (ATr). When either the main body undertaking relationship or the business connection relationship is met, the two management action records (Atr) are connected as adjacent nodes in the same control chain. The join process continues until the control relationship chain result Ctl, arranged sequentially in chronological order, is obtained.

7. The employment compliance assessment system based on big data mining according to claim 6, characterized in that: The relationship restoration module also includes a subject identification unit; The subject identification unit reads the link content in the control relationship chain result Ctl one by one, and counts the number of times each subject appears in the same control relationship chain result Ctl, the consecutive occurrence segments, and the distribution of the corresponding management links; When the same entity continuously initiates management actions in multiple management stages, or repeatedly appears as the initiator of management actions in multiple adjacent link nodes, the entity is identified as the entity that continuously controls the target employment object; then, the identified entity is written into the actual management entity result Rst.

8. The employment compliance assessment system based on big data mining according to claim 7, characterized in that: The subject verification module includes a subject comparison unit; The subject comparison unit reads the actual management subject result Rst, the contract subject set Con, the performance subject set Ful, and the registration subject set Reg corresponding to the target employment object; The actual management entity in the actual management entity result Rst is compared item by item with the contract entity in the contract entity set Con, the performance entity in the performance entity set Ful, and the registration entity in the registration entity set Reg; When the actual management entity in the actual management entity result Rst is the same as or mapped to the same entity in the corresponding set, the result is recorded as consistent; When the actual management entity in the actual management entity result Rst is different from the entity in the corresponding set, and cannot be mapped to the same entity, the result is recorded as inconsistent; Then, write the comparison results of each item into the corresponding subject verification record Vfy.

9. The employment compliance assessment system based on big data mining according to claim 8, characterized in that: The main verification module also includes a deviation determination unit; The deviation determination unit reads each subject verification record Vfy and makes a deviation determination based on the comparison results recorded in each subject verification record Vfy. When the actual management entity result Rst is consistent with the contract entity set Con, the performance entity set Ful, and the registration entity set Reg, it is determined to be without deviation; When the actual management entity's result Rst is consistent with only a portion of the set, it is judged as a partial deviation; When the actual management entity result Rst is inconsistent with the contract entity set Con, the performance entity set Ful, and the registration entity set Reg, it is judged as a significant deviation; Then write the judgment result into the main deviation result Dev.

10. The employment compliance assessment system based on big data mining according to claim 9, characterized in that: The evaluation output module includes a comprehensive evaluation unit; The comprehensive assessment unit reads the control relationship chain result Ctl and the subject deviation result Dev corresponding to the target employee, and performs employment relationship determination and compliance risk assessment based on the number of times the same subject appears in the control relationship chain result Ctl in task arrangement, shift arrangement, approval control, training distribution, assessment processing, penalty processing, remuneration confirmation, job adjustment and access control, and the deviation judgment result recorded in the subject deviation result Dev. When the same entity appears as the action initiator in at least three of the above management stages in the control relationship chain result Ctl, and the entity deviation result Dev is a significant deviation, the corresponding target employment object is determined to be a subject misalignment relationship, and the corresponding relationship determination result Jud is output. At the same time, the corresponding target employment object is assessed as a high-risk state, and the corresponding compliance assessment result Eva is output. When the same entity appears as the action initiator in at least two of the above management stages in the control relationship chain result Ctl, and the entity deviation result Dev is a partial deviation, the corresponding target user is determined to be in a subject separation relationship, and the corresponding relationship determination result Jud is output. At the same time, the corresponding target user is assessed as a medium-risk state, and the corresponding compliance assessment result Eva is output. When the same subject appears as the action initiator only in one management link in the control relationship chain result Ctl, and the subject deviation result Dev is no deviation, the corresponding target user is determined to be a subject consistency relationship, and the corresponding relationship determination result Jud is output. At the same time, the corresponding target user is assessed as a low-risk state, and the corresponding compliance assessment result Eva is output.

Images