A method and system for monitoring and tracking cooperative services among electronics manufacturing enterprises
By adopting a risk-driven dynamic monitoring strategy in the electronics manufacturing industry, and utilizing random forest models and anomaly detection technology, high-frequency and accurate status monitoring was achieved without increasing the burden on the monitored parties. This resolved the contradiction between proactive monitoring and the burden of cooperation, and improved the transparency and accuracy of monitoring decisions.
Patent Information
- Application Number
- CN202611258932.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-19
- Publication Date
- 2026-11-17
AI Technical Summary
In the electronics manufacturing industry, existing technologies using fixed-frequency monitoring methods lead to fatigue among the monitored parties, reduce the accuracy and timeliness of monitoring data, and fail to achieve high-frequency, accurate status monitoring without increasing the burden on the monitored parties.
A risk-driven dynamic monitoring strategy is adopted. By acquiring multi-source business credentials to construct a feature set, a random forest model is used to predict the probability of risk triggering, and proactive interaction requests or passive collection instructions are generated. Combined with anomaly detection and quantitative scoring, the monitoring mode can be adaptively switched, isolating the initiative of the monitoring party and the cooperation burden of the monitored party.
It enables high-frequency, low-interference status monitoring, accurately identifies potential risks, reduces data distortion and system resource waste, and improves the transparency and accuracy of supply chain collaboration and decision-making.
Smart Images

Figure CN123117777A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of enterprise management technology, and in particular to a method and system for tracking and monitoring cooperative services among enterprises in the electronics manufacturing industry. Background Technology
[0002] With the rapid development of supply chain collaboration in the electronics manufacturing industry, service follow-up monitoring is increasingly being used to identify supply anomalies and ensure production continuity. One of the core technologies for achieving efficient follow-up monitoring is a risk-driven dynamic monitoring strategy. Its goal is to build a monitoring mechanism in a complex and ever-changing business collaboration environment that can capture risks in a timely manner without interfering with the normal operations of the monitored party.
[0003] Currently, mainstream electronic manufacturing services (EMS) follow-up monitoring methods primarily employ fixed-frequency proactive interactive monitoring technology. This method first establishes a unified monitoring cycle, with the monitoring party periodically sending data requests to suppliers, subcontractors, and other monitored parties to obtain their business status information. However, this method requires the monitored parties to interrupt their current work to process data and fill out reports for each response. As the monitoring frequency increases, the burden on coordination rises sharply, leading to coordination fatigue and perfunctory responses, while the accuracy and timeliness of the monitoring data actually decrease. When further increasing the interaction frequency to shorten the anomaly detection cycle, on-site management personnel spend a significant amount of time responding to data requests, gradually viewing monitoring responses as a burden and beginning to use rough estimations instead of actual measurements, significantly degrading data accuracy. While sampling-based passive monitoring methods can reduce interactive interference, they lack proactive verification means and cannot effectively identify perfunctory behavior caused by coordination fatigue. Therefore, the biggest drawback of existing follow-up monitoring methods lies in the irreconcilable contradiction between monitoring initiative and coordination burden, making it impossible to achieve high-frequency, accurate status monitoring without increasing the burden on the monitored parties. Summary of the Invention
[0004] This invention provides a method and system for monitoring and tracking collaborative services among electronic manufacturing enterprises, in order to solve the technical problem that existing technologies cannot achieve high-frequency and accurate status monitoring without increasing the burden on the monitored party.
[0005] The present invention discloses the following technical solutions: In a first aspect, embodiments of the present invention provide a method for monitoring and tracking collaborative services among enterprises in the electronics manufacturing industry, the method comprising: Obtain multi-source business documents during the follow-up process of electronic manufacturing cooperation, extract key entities and behavioral verbs from the multi-source business documents and construct a feature set, and determine the risk trigger probability of abnormal performance behavior based on the feature set; In response to the risk trigger probability being greater than a preset probability threshold, an active interaction request is generated; in response to the risk trigger probability being less than or equal to the probability threshold, a passive collection instruction is generated. Based on the passive acquisition command, the passively acquired data is formatted and anomaly detected, and adjusted to obtain a passive monitoring dataset. Based on the proactive interaction request, the proactive interaction data is quantitatively scored to obtain a response capability evaluation value; Based on the passive monitoring dataset and the response capability assessment value, the initiative of the monitoring party and the cooperation burden of the monitored party are isolated to obtain the state decoupling result of the monitoring in the cooperation process, and to complete the follow-up monitoring of enterprise cooperation services.
[0006] Optionally, determining the risk trigger probability of abnormal performance based on the feature set includes: The feature set includes the entity word frequency of key entities and the verb word frequency of action verbs; The entity word frequency and the verb word frequency are input into a pre-trained random forest model for classification and prediction processing. The random forest model outputs the risk weights corresponding to the entity word frequency and the verb word frequency, respectively. The risk trigger probability is obtained by performing a weighted summation based on the risk weights of the entity word frequency and the verb word frequency.
[0007] Optionally, the step of generating an active interaction request in response to the risk trigger probability being greater than a preset probability threshold, and generating a passive data collection instruction in response to the risk trigger probability being less than or equal to the probability threshold, includes: The probability of risk triggering is compared with a preset probability threshold; If the probability of triggering the risk is greater than the probability threshold, an active interaction request is generated. If the risk trigger probability is less than or equal to the probability threshold, a passive acquisition instruction is generated.
[0008] Optionally, the step of formatting and anomaly detection processing on the passively acquired data based on the passive acquisition command to obtain a passive monitoring dataset includes: Based on the passive acquisition command, the passively acquired data is formatted and redundant data is removed to obtain the initial passive monitoring dataset. Anomaly detection processing is performed on the initial passive monitoring dataset to identify abnormal behaviors in the initial passive monitoring dataset, and the state corresponding to the abnormal behavior is marked as a state of coordination fatigue. Based on the aforementioned fatigue state, the monitoring frequency is adjusted, and the initial passive monitoring dataset under the aforementioned monitoring frequency is re-acquired to obtain the passive monitoring dataset.
[0009] Optionally, the step of performing anomaly detection processing on the initial passive monitoring dataset to identify abnormal behaviors in the initial passive monitoring dataset and marking the state corresponding to the abnormal behavior as a state of coordination fatigue includes: Anomaly detection processing is performed on the initial passive monitoring dataset to identify abnormal behaviors in the initial passive monitoring dataset, including repetitive operation behaviors and delayed response behaviors; The cumulative number of occurrences of the repetitive operation and the duration of the delayed response are obtained. The fatigue index is calculated by weighted summation of the number of occurrences and the duration. The state corresponding to when the fatigue index is greater than a preset index threshold is recorded as the coordination fatigue state.
[0010] Optionally, the step of quantifying and scoring the proactive interaction data based on the proactive interaction request to obtain a response capability assessment value includes: Based on the proactive interaction request, obtain the sending time and response time of the interaction information; Calculate the difference between the response time and the sending time to obtain a time difference value. Classify the active interaction data whose time difference value is greater than a preset time threshold to obtain a classification result. The classification results are mapped to obtain a classification score. The classification score and the time difference are weighted and summed to calculate the response capability assessment value.
[0011] Optionally, based on the passive monitoring dataset and the response capability assessment value, the initiative of the monitoring party and the cooperation burden of the monitored party are isolated to obtain the state decoupling result of monitoring in the cooperation process, thus completing the enterprise cooperation service follow-up monitoring, including: Obtain the anomaly frequency from the passive monitoring dataset, and calculate the product of the anomaly frequency and the response capability assessment value; The multiplicative integral is decomposed into active behavior characteristics and cooperative burden characteristics; In response to the active behavior feature being greater than a preset feature threshold, the corresponding cooperation burden feature is isolated to obtain a load index. Based on the load indicators and the proactive behavior characteristics, the state decoupling results of the monitoring in the cooperation process are generated to complete the follow-up monitoring of enterprise cooperation services.
[0012] Secondly, embodiments of the present invention provide an electronic manufacturing enterprise cooperation service follow-up monitoring system, including: The acquisition module is used to acquire multi-source business documents during the follow-up process of electronic manufacturing cooperation, extract key entities and behavioral verbs from the multi-source business documents and construct a feature set, and determine the risk trigger probability of abnormal performance behavior based on the feature set; The judgment module is used to generate an active interaction request in response to the risk trigger probability being greater than a preset probability threshold, and to generate a passive collection instruction in response to the risk trigger probability being less than or equal to the probability threshold. The detection module is used to format and detect anomalies in the passively acquired data based on the passive acquisition command, and adjust it to obtain a passive monitoring dataset. The quantification module is used to quantify and score the active interaction data based on the active interaction request to obtain a response capability evaluation value. The monitoring module is used to isolate the initiative of the monitoring party and the cooperation burden of the monitored party based on the passive monitoring dataset and the response capability evaluation value, so as to obtain the state decoupling result of the monitoring in the cooperation process and complete the follow-up monitoring of enterprise cooperation services.
[0013] Thirdly, another embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the electronic manufacturing enterprise cooperation service follow-up monitoring method as described above.
[0014] Another embodiment of the present invention also provides a computer program product, including a computer program or instructions, which, when executed by a device, implement the steps of the aforementioned electronic manufacturing enterprise cooperation service follow-up monitoring method.
[0015] Compared with the prior art, the present invention has the following beneficial effects: This invention, through the introduction of a risk-driven dynamic monitoring strategy, constructs a two-layer monitoring architecture that separates passive data collection from proactive interaction, fundamentally resolving the physical contradiction between proactive monitoring and the burden of cooperation. The passive data collection layer automatically pulls business logs at a fixed frequency, achieving high-frequency, low-interference status monitoring. The proactive interaction layer only initiates targeted confirmation requests when risks are triggered, minimizing cooperation requests. By transforming multi-source unstructured business credentials into structured risk features and using a random forest model for probability prediction, accurate identification and classification of potential cooperation risks are achieved, providing a reliable quantitative basis for dynamic switching of monitoring modes. Anomaly detection of passively collected data identifies cooperation fatigue states and adaptively adjusts the monitoring frequency, avoiding data distortion and system resource waste caused by high-frequency data collection. Dual quantitative scoring of proactive interaction response time and content objectively assesses the monitored party's true responsiveness and willingness to cooperate. Finally, by isolating proactiveness from the burden of cooperation, the positive driving behavior coupled with inefficient internal friction behavior in monitoring indicators is precisely separated, outputting state decoupling results, significantly improving the transparency, controllability, and decision-making accuracy of supply chain collaboration in the electronics manufacturing industry. Attached Figure Description
[0016] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0017] Figure 1 This is a flowchart illustrating a method for monitoring and tracking collaborative services among electronic manufacturing enterprises, as provided in an embodiment of the present invention. Figure 2 This is a schematic diagram illustrating the specific implementation process of a method for tracking and monitoring cooperative services among electronic manufacturing enterprises provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of a follow-up monitoring device for cooperative services of electronic manufacturing enterprises provided in an embodiment of the present invention. Detailed Implementation
[0018] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the invention, are intended to cover non-exclusive inclusion.
[0020] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.
[0021] See Figure 1 To address the technical problem that existing technologies struggle to achieve high-frequency, accurate status monitoring without increasing the burden on the monitored party, an embodiment of this invention provides a method for follow-up monitoring of cooperative services in the electronics manufacturing industry, comprising: S1. Obtain multi-source business documents during the follow-up process of electronic manufacturing cooperation, extract key entities and behavioral verbs from the multi-source business documents and construct a feature set, and determine the risk trigger probability of abnormal performance behavior based on the feature set.
[0022] This process involves acquiring diverse and heterogeneous business documents covering all stages of electronic manufacturing cooperation, from R&D prototyping to mass production. These documents include material procurement contracts, surface mount production schedules, printed circuit board processing drawings, first article inspection reports, and supplier delivery confirmation letters. A text processing module with a pre-built electronic manufacturing-specific dictionary performs word segmentation and part-of-speech tagging. Entity recognition rules and dependency parsing are then used to extract key entities such as equipment, materials, and processes, as well as action verbs such as procurement, mounting, welding, testing, and rework from these documents. Based on this, a structured feature set is constructed. Subsequently, a random forest model is used to classify and predict this feature set. Weighted summation and normalization are then performed by combining the co-occurrence frequency of key entities and action verbs under specific categories with a pre-set risk weight matrix. Finally, the probability of triggering risk for abnormal performance behavior is determined to quantify the risk level of delivery deviations or quality anomalies at the current cooperation follow-up stage.
[0023] S2, in response to the risk trigger probability being greater than a preset probability threshold, an active interaction request is generated; in response to the risk trigger probability being less than or equal to the probability threshold, a passive collection instruction is generated.
[0024] After calculating the risk trigger probability, the system compares this probability value with a preset probability threshold and performs adaptive traffic distribution processing in the monitoring mode based on the comparison result. When the risk trigger probability is greater than the preset probability threshold, the system determines that the current business operation is in a high-risk state and immediately generates an active interaction request. This request sends business confirmation information containing specific confirmation items to the monitored party in a strong intervention manner, such as through on-site maintenance personnel or external vendor terminals, and simultaneously triggers a secondary verification or identity review process to block potential unauthorized transfers and information leaks. Conversely, when the risk trigger probability is less than or equal to the probability threshold, the system determines that the current operation status is in a relatively safe range, and the risk indicators are not sufficient to support a strong intervention action. In this case, a passive data collection instruction is generated. This instruction triggers the passive data collection layer to pull operation logs and system login records from the partner system according to a preset collection cycle at a low sampling frequency such as 0.5Hz to 1Hz, continuously monitoring background data without disrupting the normal business rhythm of the monitored party. Through this binary decision-making mechanism based on risk probability thresholds, the system enables on-demand switching between the active interaction layer and the passive data collection layer, ensuring timely intervention in high-risk scenarios while guaranteeing low interference of monitoring behavior on the monitored party in low-risk scenarios.
[0025] S3. Based on the passive acquisition instruction, the passively acquired data is formatted and anomaly detected to obtain a passive monitoring dataset.
[0026] After generating the passive acquisition command, the system triggers the passive data acquisition layer to pull the partner's operation logs and system login records at a preset period. First, the heterogeneous data is converted into a unified structure by deduplication and field mapping formatting through a hash algorithm. Then, the isolated forest algorithm is used to detect deviations in the behavior trajectory, and the support vector machine model is used to identify repetitive operations and delayed response behaviors. After the above formatting and anomaly detection processing, a structured passive monitoring dataset with anomaly risk labels is finally obtained, which provides a data foundation for subsequent fatigue state determination and monitoring frequency adjustment.
[0027] S4. Based on the proactive interaction request, the proactive interaction data is quantitatively scored to obtain a response capability evaluation value.
[0028] After generating a proactive interaction request, the system sends business confirmation information to the monitored party through the proactive interaction layer and receives the response time and content, performing quantitative scoring from two dimensions: time and content. In the time dimension, the response time is compared to a standard response window, and an exponential decay function is used to calculate the timeliness score. In the content dimension, the response content is parsed using natural language processing, and a weighted fusion score is applied based on three sub-dimensions: completeness, accuracy, and logical consistency. Finally, the system dynamically weights and sums the time score and content score to obtain a quantified response capability assessment value, used to measure the timeliness and quality of the monitored party's response.
[0029] S5. Based on the passive monitoring dataset and the response capability evaluation value, the initiative of the monitoring party and the cooperation burden of the monitored party are isolated to obtain the state decoupling result of the monitoring in the cooperation process, and the enterprise cooperation service follow-up monitoring is completed.
[0030] The system acquires response capability assessment values and passive monitoring datasets. Manufacturing process and material flow data are extracted from the passive monitoring datasets. A support vector machine algorithm is used to construct a classification hyperplane to categorize anomalies and count their frequencies. The anomaly frequencies are multiplied by the response capability assessment values, and then decomposed into proactive behavior features and cooperation burden features using partial least squares regression. The cooperation burden features are then isolated to remove duplicate records, resulting in a load index. Finally, the independent states and proactive behavior features are merged to output a state decoupling result for the electronic manufacturing cooperation follow-up process. This result includes independent proactiveness and cooperation burden indices, and overload points and key acceleration points are marked at each stage. This enables accurate differentiation of the true cooperation intentions and execution capabilities of partners, completing the enterprise cooperation service follow-up monitoring.
[0031] Furthermore, such as Figure 2As shown, this application provides a specific implementation process for a method of monitoring and following up on cooperative services between electronic manufacturing enterprises. First, multi-source business vouchers are acquired during the electronic manufacturing cooperation follow-up process. A text processing module performs word segmentation and part-of-speech tagging on these vouchers, extracting key entities and action verbs to obtain a feature set. Then, based on the feature set, a random forest model is used for classification and prediction, calculating the risk weights corresponding to key entities and action verbs to obtain the risk trigger probability. Next, the risk trigger probability is determined. If the risk trigger probability is greater than a preset threshold, an active interaction request is generated; if the risk trigger probability is less than or equal to the preset threshold, a passive data collection instruction is generated. After obtaining the passive data collection instruction, the passive data collection layer retrieves the partner's operation logs and system login records according to a preset collection cycle. The operation logs and system login records are deduplicated and formatted to obtain the first monitoring dataset, i.e., the initial passive monitoring dataset. Then, for the initial passive monitoring dataset, a support vector machine is used for abnormal behavior detection, identifying repetitive operation behaviors and delayed response behaviors in the initial passive monitoring dataset to obtain the cooperation fatigue state. Based on the fatigue status of the cooperation, the monitoring frequency of the passive data acquisition layer is adjusted. Operation logs and system login records are then re-collected according to the adjusted monitoring frequency to obtain the second monitoring dataset, i.e., the passive monitoring dataset. Simultaneously, after receiving an active interaction request, business confirmation information is sent to the monitored party through the active interaction layer. The response time and content of the monitored party to the business confirmation information are received, and the response time and content are quantitatively scored to obtain a response capability assessment value. Finally, based on the response capability assessment value and the passive monitoring dataset, the isolation processing of initiative and cooperation burden is performed, outputting the state decoupling result for the electronic manufacturing cooperation follow-up process.
[0032] In one embodiment, determining the risk trigger probability of abnormal performance behavior based on the feature set includes steps S201 to S203, each step of which is as follows: S201, the feature set includes the entity word frequency of key entities and the verb word frequency of action verbs.
[0033] The feature set includes entity word frequencies for key entities and verb word frequencies for action verbs. Key entities refer to business objects involved in electronic manufacturing collaborations, such as equipment like pick-and-place machines and reflow ovens, materials like printed circuit boards and ceramic capacitors, and processes like surface mount technology. Their word frequencies reflect the density and level of attention these entities receive in the current business document. Action verbs represent specific technological actions performed on these objects, such as procurement, mounting, soldering, testing, and rework. Their word frequencies reflect the activity level of the business flow. After the system performs word segmentation and part-of-speech tagging on multi-source business documents through the text processing module, it counts the number of times each key entity and action verb appears in the document text, forming quantitative values for entity word frequencies and verb word frequencies. These two dimensions of word frequency data together constitute the key numerical features in the feature set, which are subsequently used as input variables by the random forest model for node splitting and classification prediction to calculate the risk weights corresponding to each entity and verb and ultimately determine the risk trigger probability.
[0034] S202, the entity word frequency and the verb word frequency are input into a pre-trained random forest model for classification and prediction processing, and the random forest model outputs the risk weights corresponding to the entity word frequency and the verb word frequency, respectively.
[0035] After constructing a feature set containing entity word frequencies and verb word frequencies, the system uses these two types of word frequency data as input feature vectors and inputs them into a pre-trained random forest model for classification and prediction. This model consists of multiple decision trees built based on different feature subsets and sample subsets. In practical applications, the sample subsets are sampled with replacement, with a sampling ratio of 0.7. Each decision tree independently evaluates node splitting for the input features. During the prediction phase, after all decision trees output classification labels, the main category is determined by majority voting. Then, the average probability of the leaf nodes under each category is taken as the confidence level. Finally, based on the probability distribution of each category output by the model and combined with a preset process risk benchmark matrix, the system outputs the risk weights corresponding to entity word frequencies (typically ranging from 0.3 to 0.7) and verb word frequencies (typically ranging from 0.2 to 0.6) respectively through table lookup or linear mapping algorithms. The weight values depend on the criticality of the entity in the process chain and the impact of the action on product yield.
[0036] S203, based on the risk weight of the entity word frequency and the risk weight of the verb word frequency, perform a weighted summation calculation to obtain the risk trigger probability.
[0037] In this process, after determining the risk weights corresponding to entity word frequencies and verb word frequencies respectively, the system performs a weighted summation operation on the risk weights of the two types of word frequencies. Specifically, the risk weight of the key entity is multiplied by its word frequency to obtain the entity risk contribution value, and the risk weight of the action verb is multiplied by its word frequency to obtain the verb risk contribution value. These two contribution values are then summed and normalized to a range of 0 to 1, thus obtaining the final risk trigger probability. This probability value comprehensively reflects the combined risk effect of both the entity object and the action in the current business operation. In specific applications, the product of a delayed action weight of 0.9 and a delivery order entity weight of 1.05 significantly increases the total probability value, providing a unified quantitative decision-making basis for subsequent judgments on whether to trigger proactive interaction requests or passive data collection instructions.
[0038] This embodiment achieves dual decoupling at the feature and model levels by inputting entity and verb word frequencies into a random forest model to independently output corresponding risk weights, and then performing weighted summation to obtain the risk trigger probability. On one hand, entity and verb word frequencies correspond to different risk weight value ranges, reflecting the differences in the criticality of the business object and the impact of the process action, respectively, ensuring that the two types of features function independently in risk assessment. On the other hand, the multi-decision tree ensemble mechanism of the random forest model ensures the stability and generalization ability of the weight output, avoiding the problem of a single decision tree overfitting specific word frequency edge values. Finally, the risk trigger probability obtained through weighted summation and normalization integrates the risk contributions from both entity and action aspects, significantly improving the accuracy and interpretability of risk warnings compared to simply statistically analyzing word frequencies or manually setting rules.
[0039] In one embodiment, in response to the risk trigger probability being greater than a preset probability threshold, an active interaction request is generated; in response to the risk trigger probability being less than or equal to the probability threshold, a passive data collection instruction is generated, including steps S301 to S303, each step of which is as follows: S301, compare the risk trigger probability with a preset probability threshold.
[0040] The system compares the calculated risk trigger probability with a preset probability threshold. In practical applications, this threshold is typically set to 0.7, and can be increased to 0.8 in highly sensitive scenarios. The specific value is set based on historical business voucher flow data and information security event samples. The comparison logic uses simple numerical judgment. S302, if the risk trigger probability is greater than the probability threshold, generate an active interaction request.
[0041] When the probability of a risk triggering exceeds a preset probability threshold, the system determines that the current business operation is in a high-risk state and immediately generates a proactive interaction request. This request sends business confirmation information containing specific confirmation items to the monitored party through the proactive interaction layer in a strong intervention manner, such as by on-site maintenance personnel or external vendor terminals. For example, it might send an interaction card requesting confirmation of the current valve opening of Cooling Pump No. 3 and whether there are any abnormal noises. Simultaneously, it triggers secondary verification, identity review, or credential tracing prompts and requires the monitored party to reply with confirmation content within a specified time window to prevent potential unauthorized transfers, information leaks, or business risks.
[0042] S303, if the risk trigger probability is less than or equal to the probability threshold, generate a passive acquisition command.
[0043] Specifically, when the probability of a risk trigger is less than or equal to a preset probability threshold, the system determines that the current business operation is in a relatively safe range, and the risk indicators are insufficient to support strong intervention. It then generates a passive data collection command. This command triggers the passive data collection layer to retrieve operation logs and system login records from the partner system at a low sampling frequency of 0.5Hz to 1Hz, according to a preset collection cycle (e.g., once every ten minutes during peak business hours and once per hour during off-peak hours). This continuous background data monitoring does not disrupt the normal business rhythm of the monitored party. In practical applications, this passive collection mechanism can reduce the real-time processing load of the onboard computing unit by 30% to 40%, while controlling the false alarm trigger rate to below 0.3 times per hour, ensuring the continuity of status monitoring.
[0044] This embodiment achieves adaptive binary switching of monitoring modes by simply comparing the risk trigger probability with a preset threshold. This allows the system to dynamically select between two differentiated monitoring paths—active interaction or passive data collection—based on the real-time risk level. When the risk probability exceeds the threshold, an active interaction request is triggered to promptly block potential risks through strong intervention. When the risk probability is below or equal to the threshold, a passive data collection command is triggered, enabling continuous monitoring in a low-frequency background mode without interfering with the normal operation of the monitored party. This threshold-based decision-making mechanism is clear in structure and highly efficient in execution. It avoids the dilemma of high-frequency interference or low-frequency missed risks caused by monitoring at a fixed single frequency in traditional solutions. While ensuring timely intervention in high-risk scenarios, it also ensures low interference with the monitored party in low-risk scenarios, achieving a preliminary decoupling of monitoring initiative and coordination burden.
[0045] In one embodiment, based on the passive acquisition instruction, the passively acquired data is formatted and anomaly detected to obtain a passive monitoring dataset, including steps S401 to S403, each step of which is as follows: S401, Based on the passive acquisition instruction, the passively acquired data is formatted to remove redundant data and obtain an initial passive monitoring dataset.
[0046] The system extracts key business fields such as operator account, operation timestamp, operation type, and terminal device identifier from operation logs and system login records, concatenates them, and calculates a fixed-length hash value using a secure hash algorithm. This hash value is then compared with historical hash values stored in the in-memory database. If duplicate hash values are found, the data is considered duplicate and removed. Subsequently, a pre-defined field mapping template is used to unify the heterogeneous log structures from different partner systems into the system's standard key-value pair format. Time fields with different expressions are converted into a standard timestamp format, and missing non-critical fields are added. After deduplication and format alignment, a unified and non-redundant initial passive monitoring dataset is obtained, providing a clean data foundation for subsequent anomaly detection and status assessment.
[0047] S402, perform anomaly detection processing on the initial passive monitoring dataset, identify abnormal behaviors in the initial passive monitoring dataset, and mark the state corresponding to the abnormal behavior as a state of coordination fatigue.
[0048] The system employs a combination of statistical operation frequency and response time difference to form behavioral characteristics. These characteristics are then input into a pre-trained support vector machine model to construct a decision boundary that distinguishes between normal and abnormal behaviors. This identifies repetitive operations, such as issuing multiple invalid commands to the same control valve within a short period, and delayed response behaviors, such as response times exceeding 2.5 times the historical average after an alarm. These two types of abnormal behaviors are used as the basis for judgment. Within the evaluation period, the system performs a weighted sum of the frequency of identified repetitive operations and the number of delayed response behaviors. When the weighted result reaches a preset fatigue judgment threshold, the state corresponding to this abnormal behavior is marked as a state of coordination fatigue. This state specifically refers to the decline in human-machine coordination ability caused by physiological and attentional deterioration in the monitored entity, and is used for adaptive adjustment of subsequent monitoring frequency triggers.
[0049] S403, based on the aforementioned fatigue state, adjust the monitoring frequency, re-acquire the initial passive monitoring dataset at the aforementioned monitoring frequency, and obtain the passive monitoring dataset.
[0050] After determining the fatigue state of the system, the monitoring frequency is dynamically adjusted according to the level of fatigue. In practical applications, when the fatigue state is mild, the monitoring frequency adjustment factor is set to 80%, which is a 20% reduction from the original sampling frequency. When the fatigue state is severe, the adjustment factor is set to 30%, and an intermittent sleep mechanism is even triggered, extending the sampling interval from once per minute to once every five minutes. The system re-collects operation logs and system login records according to the adjusted target monitoring frequency, performs deduplication and formatting processing again, and finally obtains a passive monitoring dataset. This dataset has a lower sampling density than the initial high-frequency data, but by actively reducing the system load, it ensures the integrity and reliability of the collected data, providing high-quality data support for subsequent fatigue isolation processing.
[0051] This embodiment forms a complete data quality optimization and dynamic feedback loop by sequentially performing formatting and deduplication, abnormal behavior identification, and adaptive adjustment of monitoring frequency on passively collected data. First, formatting removes redundant data, eliminating duplicate records caused by heterogeneous log structure differences and network retransmissions, providing a clean data foundation for subsequent analysis. Second, abnormal behavior identification identifies repetitive operations and delayed responses, marking a state of cooperation fatigue, enabling the system to quantify the actual cooperation burden level of the monitored party. Finally, the monitoring frequency is dynamically reduced based on the cooperation fatigue state, avoiding further burdening the other party with high-frequency collection under fatigue conditions. This ensures the sustainability of data collection while maintaining the integrity and reliability of the monitoring data. This progressive processing mechanism effectively resolves the rigid contradiction between collection intensity and cooperation burden in traditional fixed-frequency monitoring, significantly improving the system's adaptability and data collection quality under passive monitoring mode.
[0052] In one embodiment, anomaly detection processing is performed on the initial passive monitoring dataset to identify abnormal behaviors in the initial passive monitoring dataset, and the state corresponding to the abnormal behavior is marked as a fatigue state. This includes steps S501 to S503, each of which is as follows: S501, perform anomaly detection processing on the initial passive monitoring dataset to identify abnormal behaviors in the initial passive monitoring dataset, including repetitive operation behaviors and delayed response behaviors.
[0053] The system counts the frequency of the same type of operation command within a preset sliding time window and the time interval between two consecutive identical operations. When the frequency exceeds the normal operation threshold and the time interval is less than the thinking and execution time required for normal business logic, it is quantified and marked as a repetitive operation feature value to capture mechanical redundancy caused by inattention. Simultaneously, the system extracts the difference between the system event trigger timestamp and the operator's first effective intervention timestamp. When this time difference is significantly greater than the average level of historical normal response time or exceeds the maximum allowable delay time specified in safety procedures, it is converted into a delayed response feature value. These two types of feature values together constitute the input of the support vector machine model for subsequent determination of fatigue state.
[0054] S502, obtain the cumulative number of occurrences of the repetitive operation and the duration of the delayed response, and calculate the fatigue index by weighted summation of the number of occurrences and the duration.
[0055] After identifying repetitive operations and delayed response behaviors, the cumulative number of repetitive operations within a set evaluation period is counted. Number of delayed response behaviors (A single action whose duration exceeds the normal response threshold is counted as one instance), and a weighted sum is calculated using a preset weighting coefficient. The specific calculation formula is as follows:
[0056] In the formula, and and The value can be adjusted between 0 and 1 according to the on-site working conditions and must meet the following requirements. + =1, in typical configuration Take 0.6, The value is set to 0.4. The system uses the weighted summation value F as the fatigue index. This index comprehensively reflects the overall severity of abnormal behavior. When F reaches the preset fatigue judgment threshold, the system determines that the current operator is in a state of cooperative fatigue, providing a quantitative basis for the adaptive adjustment of the subsequent monitoring frequency.
[0057] S503, the state corresponding to when the fatigue index is greater than the preset index threshold is recorded as the coordination fatigue state.
[0058] The calculated fatigue index F is compared with a preset threshold. When F exceeds the threshold, the operator's status within the current evaluation period is recorded as a state of coordination fatigue. This threshold is typically preset based on historical operational data and safety procedures; in practical applications, a typical value is 8, meaning that the accumulated abnormal behavior within the evaluation window has exceeded the upper limit of the normal fluctuation range. Once a state of coordination fatigue is determined, it indicates that the operator's human-machine coordination ability has declined due to decreased physiological and attentional abilities. The system will then trigger subsequent monitoring frequency adjustment strategies, such as reducing the collection frequency or extending the collection interval, to alleviate the coordination burden on the monitored party and ensure the sustainability of data collection.
[0059] This embodiment refines abnormal behavior into two quantifiable indicators: repetitive operations and delayed responses. The cumulative number of occurrences and duration of each indicator are statistically calculated, and then weighted and summed to form a fatigue index. This index is then compared with a preset threshold to determine the cooperation fatigue state, achieving an objective quantitative assessment of the monitored party's cooperation burden level. Compared to schemes relying solely on a single behavioral indicator or subjective judgment, this mechanism effectively avoids misjudgments caused by occasional anomalies, ensures the robustness and repeatability of status marking, and provides a reliable quantitative triggering basis for the adaptive adjustment of subsequent monitoring frequency.
[0060] In one embodiment, based on the proactive interaction request, the proactive interaction data is quantitatively scored to obtain a response capability assessment value, including steps S601 to S603, each step of which is as follows: S601, Based on the active interaction request, obtain the sending time and response time of the interaction information.
[0061] Based on the generated proactive interaction request, a business confirmation message is sent to the monitored party through the proactive interaction layer, and the sending time of this message is recorded. When the monitored party submits feedback data through the terminal, the system synchronously records the response time when the feedback data is fully received. Both timestamps are uniformly marked by the system's internal clock. The sending time is taken from the time when the proactive interaction layer sends the business confirmation message, and the response time is taken from the time when the system fully receives the feedback data submitted by the monitored party. The time difference between the two, i.e., the response time minus the sending time, serves as the basis for subsequent timeliness scoring, used to quantitatively evaluate the timeliness of the monitored party's response.
[0062] S602, calculate the difference between the response time and the sending time to obtain a time difference, and classify the active interaction data whose time difference is greater than a preset time threshold to obtain a classification result.
[0063] The system has a pre-defined standard response time window. When the time difference exceeds the maximum tolerable time threshold, the system resets the time dimension score to zero and marks the interaction as timed out. For data where the time difference does not exceed the maximum tolerable time but exceeds the optimal response window, the system uses an exponential decay function to calculate the timeliness score and classifies it into different categories such as "fast response," "normal response," and "delayed response" based on the score's range. The larger the time difference and the lower the timeliness score, the more negative the classification result leans. These classification results, together with the classification score of the response content, form the basis of the quantitative score, providing a time dimension basis for calculating the final response capability assessment value.
[0064] S603, map the classification result to obtain a classification score, and calculate the response capability assessment value by weighted summation of the classification score and the time difference.
[0065] The system, after obtaining the classification results corresponding to the time difference, converts different categories into corresponding classification scores using preset mapping rules. Then, it performs a weighted summation of these classification scores and the time difference to obtain a response capability assessment value. Specifically, the time difference itself serves as a quantitative indicator of timeliness, while the classification score reflects the corresponding level of timeliness. The two are integrated through a dynamic weight allocation strategy. That is, the system adjusts the weight ratio according to the urgency of the current business confirmation information. For security-related confirmation information, the time dimension is emphasized; for routine inspection information, the content dimension is emphasized. After weighted summation and normalization, a comprehensive quantitative score, the response capability assessment value, is finally output. This value simultaneously reflects the response speed level and actual delay of the monitored party, and is used for subsequent proactive assessment and isolation of cooperation burden.
[0066] This embodiment obtains a classification score by classifying and mapping the time difference between the response time and the sending time, and then weights and sums it with the time difference itself. This allows the response capability assessment value to simultaneously integrate both actual response speed and timeliness level evaluation information. It retains the quantitative accuracy of continuous values and enhances the stability of the assessment through classification and mapping, avoiding the defects of single numerical indicators being easily affected by extreme values. This provides a more comprehensive and reliable quantitative basis for subsequent proactive assessment and isolation of cooperation burden.
[0067] In one embodiment, based on the passive monitoring dataset and the response capability assessment value, the initiative of the monitoring party and the cooperation burden of the monitored party are isolated to obtain the state decoupling result of the monitoring in the cooperation process, and the enterprise cooperation service follow-up monitoring is completed, including steps S701 to S704, each step as follows: S701, Obtain the anomaly frequency from the passive monitoring dataset, and calculate the product of the anomaly frequency and the response capability assessment value.
[0068] The system extracts manufacturing process and material flow data from the passive monitoring dataset. It then uses a support vector machine algorithm to construct a classification hyperplane to categorize anomalies and count the number of anomaly categories, obtaining the anomaly frequency. The anomaly frequency reflects the overall occurrence frequency of abnormal behaviors such as repetitive operations and delayed responses in the partner's business operations. Subsequently, the anomaly frequency is multiplied by the response capability assessment value calculated in the previous steps to obtain the product. This product serves as an intermediate variable for subsequent isolation processing. Its value comprehensively reflects the correlation strength between anomaly occurrence density and the monitored party's proactive response capability. When the anomaly frequency is high and the response capability assessment value is low, the product result is large, indicating a coupled state of low response capability accompanied by high anomaly frequency. Partial least squares regression is needed to further decompose this product result into independent proactive behavior features and cooperation burden features to achieve accurate separation of the two states.
[0069] S702, the multiplicative integral is decomposed into active behavior characteristics and cooperative burden characteristics.
[0070] The system maps interactive behaviors in the passively monitored dataset into multi-dimensional feature vectors, covering dimensions such as response time, message round trips, and document modification frequency. A behavioral feature matrix X is constructed, and a target vector Y is built by combining the product results. Principal components and corresponding load vectors in the feature space are extracted through iterative calculation, and the Pearson correlation coefficient between each principal component and response capability is calculated. When the correlation coefficient is greater than or equal to 0.6, the original feature component corresponding to that principal component is classified as an active behavioral feature, reflecting the proactive behavior of the partner in advancing the project, such as early warning of supply chain risks. When the correlation coefficient is less than -0.4 and the absolute value of the weight of the task stagnation time or resource consumption increase in the load vector is greater than 0.5, the relevant feature component is classified as a cooperation burden feature, reflecting the drag on the main task caused by frequent unnecessary interactions. Through this mathematical transformation, the system effectively separates the two coupled behavioral contributions in the product result, providing independent feature components for subsequent isolation processing.
[0071] S703, in response to the active behavior feature being greater than a preset feature threshold, the corresponding cooperation burden feature is isolated to obtain a load index.
[0072] The process involves multiplying and integrating to obtain proactive behavior characteristics and cooperation burden characteristics. The quantified values of the proactive behavior characteristics are then compared to preset thresholds. When the proactive behavior characteristic exceeds this threshold, the collaborator is deemed to have exhibited significant positive progress in the current follow-up process. In this case, the system performs isolation processing on the corresponding cooperation burden characteristics. This involves removing invalid communication records and redundant data requests arising from proactive behavior, such as unavoidable coordination confirmation records during normal progress, while retaining only the burden components that are purely unnecessary interactions. This process yields a load index. This load index is independent of the proactive behavior characteristics and accurately reflects the true degree of cooperation burden caused by inefficient internal friction or redundant processes in collaboration, providing a clean quantitative basis for the subsequent output of state decoupling results.
[0073] S704, Based on the load index and the active behavior characteristics, generate the state decoupling result of the monitoring in the cooperation process, and complete the follow-up monitoring of enterprise cooperation services.
[0074] After obtaining the isolated load indicators and proactive behavior characteristics, the system quantifies them into independent cooperation burden indices Lc and Pa initiative indices, both ranging from 0 to 1. These indices can be read separately without interference, and are then combined to generate a state decoupling result for the electronic manufacturing cooperation follow-up process. This decoupling result not only includes independent quantified values for the cooperation burden index and initiative index but also includes a detailed state distribution map. It marks overload points where the cooperation burden index exceeds 0.7 and key acceleration points where the initiative index exceeds 0.8 at each stage, from sample confirmation and material kitting to trial production ramp-up and mass production delivery. This facilitates quick identification of problem areas by the follow-up personnel. Finally, the system outputs this state decoupling result as the final output for monitoring the cooperation process, achieving accurate differentiation between the true cooperation intentions and execution capabilities of the partners. This avoids misjudgments caused by a single monitoring dimension, thus completing the overall process of enterprise cooperation service follow-up monitoring.
[0075] This embodiment decomposes the anomaly frequency into independent proactive behavior characteristics and cooperation burden characteristics by multiplying it by the response capability assessment value. The corresponding cooperation burden characteristics are isolated only when the proactive behavior characteristics exceed a threshold, achieving accurate differentiation between proactive actions and ineffective internal friction burdens. This mechanism effectively avoids misjudging necessary communication in normal collaboration as a burden, while preserving the authenticity of burden characteristics in low-proactive scenarios. The final state decoupling result reflects both the collaborator's true willingness to collaborate and its effective execution capability, providing objective data support for managers to accurately locate process bottlenecks and optimize collaboration strategies.
[0076] Based on the same inventive concept, this application also provides a system for implementing the above-mentioned follow-up monitoring system for cooperative services of electronic manufacturing enterprises. The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the follow-up monitoring device for cooperative services of electronic manufacturing enterprises provided below can be found in the limitations of the follow-up monitoring method for cooperative services of electronic manufacturing enterprises described above, and will not be repeated here.
[0077] In one exemplary embodiment, such as Figure 3 As shown, an electronic manufacturing enterprise cooperation service follow-up monitoring system is provided, including: The acquisition module 801 is used to acquire multi-source business documents in the electronic manufacturing cooperation follow-up process, extract key entities and behavioral verbs from the multi-source business documents and construct a feature set, and determine the risk trigger probability of abnormal performance behavior based on the feature set. The judgment module 802 is used to generate an active interaction request in response to the risk trigger probability being greater than a preset probability threshold, and to generate a passive collection instruction in response to the risk trigger probability being less than or equal to the probability threshold. The detection module 803 is used to perform formatting and anomaly detection processing on the passively acquired data based on the passive acquisition command, and adjust it to obtain a passive monitoring dataset. The quantization module 804 is used to quantify and score the active interaction data based on the active interaction request to obtain a response capability evaluation value. The monitoring module 805 is used to isolate the initiative of the monitoring party and the cooperation burden of the monitored party based on the passive monitoring dataset and the response capability evaluation value, so as to obtain the state decoupling result of the monitoring in the cooperation process and complete the follow-up monitoring of enterprise cooperation services.
[0078] In one embodiment, a computer program product is provided, including a computer program or instructions, which, when executed by a device, implement the steps of the aforementioned electronic manufacturing enterprise cooperation service follow-up monitoring method.
[0079] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.
[0080] This technical solution separates the passive data acquisition layer from the active interaction layer at the system level and integrates the intelligent discrimination mechanism of random forest model and support vector machine to construct a state decoupling and adaptive monitoring system specifically for follow-up on electronic manufacturing collaboration. Its core benefit lies in breaking through the limitations of existing technologies that rely solely on a single monitoring frequency or fixed interaction channel. It achieves a dynamic balance between high-frequency state acquisition and low cooperation burden. Users can achieve seamless monitoring through background log retrieval in low-risk stages and obtain accurate feedback through targeted business confirmation in high-risk stages, thus providing clear state judgment criteria for supply chain collaboration and achieving a leap from one-way pressure to intelligent diversion. Simultaneously, this solution fully leverages the respective advantages of text processing in voucher parsing, machine learning in risk identification, and dual-channel acquisition in behavior assessment, resolving the irreconcilable contradiction between proactive monitoring and the cooperation burden of the monitored party in traditional solutions.
[0081] For the device embodiments, since they basically correspond to the method embodiments, the relevant details can be found in the descriptions of the method embodiments. The device embodiments described above are merely illustrative; components described as separate parts may or may not be physically separate, and components shown as units may or may not be physical units, meaning they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without any inventive effort.
[0082] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.
Claims
1. A method for tracking and monitoring cooperative services with electronics manufacturing enterprises, characterized in that, include: Obtain multi-source business documents during the follow-up process of electronic manufacturing cooperation, extract key entities and behavioral verbs from the multi-source business documents and construct a feature set, and determine the risk trigger probability of abnormal performance behavior based on the feature set; In response to the risk trigger probability being greater than a preset probability threshold, an active interaction request is generated; in response to the risk trigger probability being less than or equal to the probability threshold, a passive collection instruction is generated. Based on the passive acquisition command, the passively acquired data is formatted and anomaly detected, and adjusted to obtain a passive monitoring dataset; Based on the proactive interaction request, the proactive interaction data is quantitatively scored to obtain a response capability evaluation value; Based on the passive monitoring dataset and the response capability assessment value, the initiative of the monitoring party and the cooperation burden of the monitored party are isolated to obtain the state decoupling result of the monitoring in the cooperation process, and to complete the follow-up monitoring of enterprise cooperation services.
2. The method for follow-up monitoring of cooperative services for electronic manufacturing enterprises as described in claim 1, characterized in that, The determination of the risk trigger probability regarding abnormal performance based on the feature set includes: The feature set includes the entity word frequency of key entities and the verb word frequency of action verbs; The entity word frequency and the verb word frequency are input into a pre-trained random forest model for classification and prediction processing. The random forest model outputs the risk weights corresponding to the entity word frequency and the verb word frequency, respectively. The risk trigger probability is obtained by performing a weighted summation based on the risk weights of the entity word frequency and the verb word frequency.
3. The method for follow-up monitoring of cooperative services for electronic manufacturing enterprises as described in claim 1, characterized in that, The method of generating an active interaction request in response to a risk trigger probability greater than a preset probability threshold, and generating a passive data collection instruction in response to a risk trigger probability less than or equal to the probability threshold, includes: The probability of risk triggering is compared with a preset probability threshold; If the probability of triggering the risk is greater than the probability threshold, an active interaction request is generated. If the probability of triggering the risk is less than or equal to the probability threshold, a passive acquisition instruction is generated.
4. The method for follow-up monitoring of cooperative services for electronic manufacturing enterprises as described in claim 1, characterized in that, The process of formatting and anomaly detection of the passively acquired data based on the passive acquisition command, and adjusting it to obtain a passive monitoring dataset, includes: Based on the passive acquisition command, the passively acquired data is formatted to remove redundant data and obtain an initial passive monitoring dataset. Anomaly detection processing is performed on the initial passive monitoring dataset to identify abnormal behaviors in the initial passive monitoring dataset, and the state corresponding to the abnormal behavior is marked as a state of coordination fatigue. Based on the aforementioned fatigue state, the monitoring frequency is adjusted, and the initial passive monitoring dataset under the aforementioned monitoring frequency is re-acquired to obtain the passive monitoring dataset.
5. The method for follow-up monitoring of cooperative services for electronic manufacturing enterprises as described in claim 4, characterized in that, The step of performing anomaly detection processing on the initial passive monitoring dataset, identifying abnormal behaviors in the initial passive monitoring dataset, and marking the state corresponding to the abnormal behavior as a state of fatigue includes: Anomaly detection processing is performed on the initial passive monitoring dataset to identify abnormal behaviors in the initial passive monitoring dataset, including repetitive operation behaviors and delayed response behaviors; The cumulative number of occurrences of the repetitive operation and the duration of the delayed response are obtained. The fatigue index is calculated by weighted summation of the number of occurrences and the duration. The state corresponding to when the fatigue index is greater than the preset index threshold is recorded as the coordinated fatigue state.
6. The method for follow-up monitoring of cooperative services for electronic manufacturing enterprises as described in claim 1, characterized in that, The step of quantifying and scoring the proactive interaction data based on the proactive interaction request to obtain a response capability assessment value includes: Based on the proactive interaction request, obtain the sending time and response time of the interaction information; Calculate the difference between the response time and the sending time to obtain a time difference value. Classify the active interaction data whose time difference value is greater than a preset time threshold to obtain a classification result. The classification results are mapped to obtain a classification score. The classification score and the time difference are weighted and summed to calculate the response capability assessment value.
7. The method for follow-up monitoring of cooperative services for electronic manufacturing enterprises as described in claim 1, characterized in that, Based on the passive monitoring dataset and the response capability assessment value, the initiative of the monitoring party and the cooperation burden of the monitored party are isolated to obtain the state decoupling result of monitoring in the cooperation process, thus completing the enterprise cooperation service follow-up monitoring, including: Obtain the anomaly frequency from the passive monitoring dataset, and calculate the product of the anomaly frequency and the response capability assessment value; The multiplicative integral is decomposed into active behavior characteristics and cooperative burden characteristics; In response to the active behavior feature being greater than a preset feature threshold, the corresponding cooperation burden feature is isolated to obtain a load index. Based on the load indicators and the proactive behavior characteristics, the state decoupling results of the monitoring in the cooperation process are generated to complete the follow-up monitoring of enterprise cooperation services.
8. A collaborative service follow-up monitoring system for electronics manufacturing enterprises, characterized in that, The system includes: The acquisition module is used to acquire multi-source business documents during the follow-up process of electronic manufacturing cooperation, extract key entities and behavioral verbs from the multi-source business documents and construct a feature set, and determine the risk trigger probability of abnormal performance behavior based on the feature set; The judgment module is used to generate an active interaction request in response to the risk trigger probability being greater than a preset probability threshold, and to generate a passive collection instruction in response to the risk trigger probability being less than or equal to the probability threshold. The detection module is used to format and detect anomalies in the passively acquired data based on the passive acquisition command, and adjust it to obtain a passive monitoring dataset. The quantification module is used to quantify and score the active interaction data based on the active interaction request to obtain a response capability evaluation value. The monitoring module is used to isolate the initiative of the monitoring party and the cooperation burden of the monitored party based on the passive monitoring dataset and the response capability evaluation value, so as to obtain the state decoupling result of the monitoring in the cooperation process and complete the follow-up monitoring of enterprise cooperation services.
9. A computer-readable storage medium, characterized in that, The storage medium stores a computer program or instructions, which, when executed by a communication device, implement the electronic manufacturing enterprise cooperation service follow-up monitoring method as described in any one of claims 1-7.
10. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by the communication device, the electronic manufacturing enterprise cooperation service follow-up monitoring method as described in any one of claims 1-7 is implemented.
