Target personnel dynamic risk early warning method and device

Abstract

The invention relates to the field of personnel management, provides a target personnel dynamic risk early warning method and device, and realizes target personnel risk early warning by combining edge calculation and artificial intelligence in order to facilitate the management of field staff. Firstly, data processing is localized by using an edge computing technology, the risk of data leakage is reduced, behavior pattern data is identified by using a behavior pattern identification model, after abnormal data is identified, data re-checking is performed by using an abnormal data re-checking model, and the data is encrypted and uploaded to the cloud only when the data is abnormal, so that the monitoring efficiency and accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the field of personnel management, and specifically to a method and device for dynamically warning of risks of target personnel. Background Art

[0002] In modern enterprise management, the management and warning of employees' working status is an important task. Effective management of employees' working status can not only improve the overall efficiency of the team, but also effectively prevent potential risks, such as work slack, violations, etc. To achieve this goal, a variety of technical methods and systems have emerged, such as intelligent attendance systems, work computer monitoring software, project management systems, etc.

[0003] However, in the process of dynamically warning of risks for field staff, since the working locations of field staff are usually not in a fixed office area, but are scattered in various corners of the city, and may even need to travel frequently, it is difficult to accurately record their attendance and movements through traditional intelligent attendance systems (such as face recognition, fingerprint recognition). For work computer monitoring software, field staff may not always be working in front of the computer, so these software also have limitations in collecting their working status data. Summary of the Invention

[0004] In order to facilitate the management of field staff, the present application provides a method and device for dynamically warning of risks of target personnel.

[0005] The technical solution adopted by the present invention to solve the above problems is: A method for dynamically warning of risks of target personnel, comprising: Step 1: Deploy mobile edge computing devices; Step 2: Collect behavior pattern data and external data of target personnel based on the edge computing devices, wherein the behavior pattern data includes location data, behavior data and work-related data of the target personnel; the external data includes at least employee task data, customer visit history and expected duration, traffic data and meteorological data; Step 3: Create and train a behavior pattern recognition model based on the behavior pattern data, and create and train an abnormal data review model based on the behavior pattern data and external data; Step 4: Real-time collect the behavior pattern data of the target personnel, and use the behavior pattern recognition model for recognition. When it is detected that the current behavior pattern data is abnormal data, based on the abnormal data and the external data corresponding to the abnormal data time period, use the abnormal data review model to confirm the abnormal data. If it is confirmed as abnormal data, upload the relevant information of the abnormal data to the central server.

[0006] Further, the edge computing devices include smart phones and tablets.

[0007] Further, step 4 determines the data collection frequency according to whether the target person is during working hours.

[0008] Further, it is determined whether it is during working hours according to the task content of the target person and the behavior of the target person during the current time period.

[0009] Further, it further includes step 5: aggregating and analyzing the information related to the abnormal data uploaded to the central server, and generating a risk distribution map and / or a risk report according to the results of the aggregation and analysis.

[0010] Further, step 5 further includes: warning the management personnel according to the results of the aggregation and analysis.

[0011] The dynamic risk warning device for the target person includes: an edge computing device and a central server. The edge computing device is provided with: A data collection unit: used to collect the behavior pattern data of the target person and external data; A model construction unit: creating and training a behavior pattern recognition model based on the behavior pattern data, and creating and training an abnormal data review model based on the behavior pattern data and external data; An abnormal data recognition unit: identifying whether the current behavior pattern data of the target person is abnormal data based on the behavior pattern recognition model; An abnormal data confirmation unit: confirming the abnormal data by using the abnormal data review model based on the abnormal data and the external data corresponding to the abnormal data time period. If it is confirmed as abnormal data, the information related to the abnormal data is uploaded to the central server.

[0012] Further, the edge computing device further includes a working time determination unit, which is used to determine whether it is during working hours according to the task content of the target person and the behavior of the target person during the current time period; the data collection unit determines the data collection frequency according to whether it is during working hours.

[0013] Further, the central server includes an information aggregation unit: used to aggregate abnormal information; an alarm unit: used to warn the management personnel according to the results of the aggregation and analysis.

[0014] Further, the central server further includes a storage unit for storing data based on a classified storage strategy; an index unit for querying the stored data.

[0015] The beneficial effects of the present invention compared with the prior art are: combining edge computing and artificial intelligence to realize the risk warning of the target person. First, using edge computing technology for local data processing to reduce the risk of data leakage. Using the behavior pattern recognition model to identify the behavior pattern data, after identifying the abnormal data, using the abnormal data review model for data review, and only encrypting and uploading to the cloud when it is abnormal, which improves the monitoring efficiency and accuracy.

[0016] Edge computing technology allows edge nodes to be deployed on multiple work areas or critical devices, forming a distributed data processing network, improving the flexibility and scalability of the system, and being able to easily meet the dynamic risk warning requirements of a large number of employees.

[0017] By establishing a behavior pattern recognition model and an abnormal data review model, learning and modeling the daily behaviors of employees can more accurately identify abnormal behaviors. Description of the Drawings

[0018] Figure 1 It is a flow chart of the dynamic risk warning method for target personnel; Figure 2 It is a schematic structural diagram of the dynamic risk warning device for target personnel. Detailed Embodiments

[0019] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below in conjunction with embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0020] As Figure 1 shown, the dynamic risk warning method for target personnel includes: Step 1: Deploy mobile edge computing devices, such as smartphones, tablets or dedicated mobile devices. Monitoring target personnel based on mobile devices improves the flexibility of monitoring.

[0021] Step 2: Collect the behavior pattern data and external data of target personnel based on the edge computing device.

[0022] The behavior pattern data includes location data, behavior data and work-related data. The location data can be collected through a GPS module. The behavior data includes moving speed, staying time, etc., which can be collected through an accelerometer, gyroscope, etc. The work-related data is obtained by filling in forms such as customer visit records and sales orders.

[0023] After the behavior pattern data is collected, it is first preprocessed on the edge computing device, including duplicate removal, noise reduction, format conversion, etc., to improve the accuracy and efficiency of subsequent analysis.

[0024] Since sensors may resend data due to network latency, device failures, etc., in the data preprocessing stage, the system first needs to deduplicate the received raw data. This is achieved by comparing fields such as the timestamp, sensor ID, and data content of the data to ensure that each piece of data is unique. Sensor data is subject to various interferences during transmission, such as electromagnetic noise, device aging, etc., resulting in data fluctuations or anomalies. To remove these noises, the system uses filtering algorithms, such as mean filtering, median filtering, etc., to smooth the raw data, thereby obtaining more accurate data.

[0025] Since different sensors use different data formats and encoding methods, in the data preprocessing stage, the system needs to convert and unify these data. This includes converting the data from encoding methods such as binary and hexadecimal to text or numerical formats that are easy to process, and converting the data formats of different sensors to a unified format, such as JSON, CSV, etc. Doing so can not only simplify the subsequent data processing flow but also improve the readability and analyzability of the data.

[0026] Data that does not conform to normal conditions due to device failures, incorrect operations, etc. If this data is used for subsequent analysis without being processed, it will lead to deviations or errors in the analysis results. Therefore, in the data preprocessing stage, the system needs to set reasonable thresholds and filtering rules to detect anomalies in the raw data. This can be achieved through statistical methods (such as mean, variance, standard deviation, etc.), machine learning algorithms (such as clustering, classification, etc.), or domain knowledge, etc. Once abnormal data is detected, the system marks it as invalid data and excludes it from subsequent analysis. Then, the data is scaled or transformed according to certain rules to ensure that data from different sources are consistent in dimension and value range. This is crucial for subsequent data processing and analysis. For example, if the value ranges of data from different sensors vary greatly, there will be deviations when comparing or analyzing the data. Therefore, in the data preprocessing stage, the system needs to standardize the raw data, such as scaling the data to the 0-1 range, performing Z-score standardization, etc. This can ensure that data from different sensors can participate in calculations and comparisons fairly in subsequent analysis.

[0027] In addition to the above operations, the data preprocessing stage also needs to establish a data quality control and monitoring mechanism. This includes regularly calibrating and maintaining the sensors to ensure the accuracy and reliability of the data; establishing data quality assessment indicators and a monitoring system to monitor and evaluate the integrity, accuracy, consistency, etc. of the data in real time; and setting data quality thresholds and an alarm mechanism to trigger an alarm and take corresponding processing measures once the data quality is below the set threshold.

[0028] External data is used to assist in determining whether the abnormal data initially identified by the behavior pattern recognition model is a real abnormality or a false abnormality. In this embodiment, the external data comes from the enterprise's internal business systems and external data platforms. Enterprise internal business systems: including a task assignment system (obtaining daily task details and planned time arrangements of employees), a customer relationship management system (understanding customer visit history and expected duration, etc.), and a project progress management system (grasping the urgency and execution standards of project tasks, etc.). External data platforms: such as a traffic data platform (obtaining information such as road congestion conditions and public transportation operation times), a meteorological data platform (understanding the impact of weather conditions on outdoor work, such as work delays caused by heavy rain), and an industry data platform (providing reference data such as industry-standard working hours and behavior patterns).

[0029] For example, if the abnormal data is that a field staff member stays in a certain area for too long, the system will compare the time period of this abnormal data with the task plan time in the employee task assignment system, and at the same time integrate the traffic condition data and weather data in the same time period in this area. The integrated data forms a comprehensive data set containing the employee's behavior data and its related environmental factors, providing a comprehensive basis for subsequent review and analysis.

[0030] Step 3: Create a behavior pattern recognition model based on the behavior pattern data and train it, and create an abnormal data review model based on the behavior pattern data and external data and train it.

[0031] Create a behavior pattern recognition model based on machine learning algorithms. Through training, the behavior pattern recognition model can learn the normal characteristics of the target person's behavior and can automatically identify abnormal data in subsequent analysis. In addition, a variety of optimization strategies are also adopted to improve the performance of the algorithm model. For example, the diversity of training data is increased through data augmentation techniques to improve the generalization ability of the model; techniques such as regularization and Dropout are used to prevent the model from overfitting and improve the stability of the model; and techniques such as hyperparameter tuning are used to find the optimal model parameters to further improve the accuracy of the model.

[0032] Create an abnormal data review model based on machine learning algorithms such as support vector machine (SVM) and random forest, and train it based on normal data, abnormal data, and the corresponding external data. After training, the model can complete the confirmation process of abnormal data.

[0033] For example, if external data indicates severe traffic congestion in the area during the same time period and the employee's work behavior shows an overly long stay time, the model will comprehensively consider these factors to determine whether it is a reasonable situation. If it is a reasonable situation, the abnormal data is determined to be a misjudgment, and the "abnormal data" label is corrected; conversely, if the external data cannot reasonably explain the abnormal behavior, the model confirms the validity of the abnormal data and generates a detailed description of the abnormal situation.

[0034] Step 4: Real-time collect the behavior pattern data of the target person and use the behavior pattern recognition model for recognition. When the current behavior pattern data is detected as abnormal data, based on the abnormal data and the external data corresponding to the abnormal data time period, use the abnormal data review model to confirm the abnormal data. If it is confirmed as abnormal data, upload the relevant information of the abnormal data to the central server.

[0035] After the behavior pattern recognition model identifies abnormal data, the system will automatically generate a data record containing special tags. The tag content covers key information such as the type of abnormal data (such as location deviation, abnormal behavior, etc.), the occurrence time, and the unique identification code of the corresponding employee. This step ensures that in the subsequent review process, the data segment and the corresponding external data that need to be reviewed can be quickly and accurately located.

[0036] Furthermore, to avoid affecting the normal life of the target person, in this embodiment, the data collection frequency is determined according to whether the target person is at work. If at work, the data collection frequency is increased; if not at work, the data collection frequency is decreased. In this embodiment, it is judged whether it is at work according to the task content of the target person and the behavior of the target person in the current time period. It is also possible to dynamically adjust the data collection frequency according to the matching degree between the behavior and the task content in the current time period.

[0037] To facilitate the management of all target persons, after confirming the abnormality, the abnormal information is uploaded to the central server for summary analysis. And according to the summary analysis, by analyzing the occurrence frequency, duration, associated factors, etc. of the abnormal behavior, potential risk laws and trends are revealed, and a risk distribution map and / or a risk report are generated according to the analysis results, summarizing the occurrence situation, handling results, and improvement measures of the abnormal behavior, etc., so that the management personnel can understand the operation effect of the monitoring mechanism and adjust and optimize the working environment and monitoring strategy accordingly. During the data upload process, encrypted upload is adopted to avoid the risk of privacy leakage caused by comprehensive monitoring.

[0038] To facilitate the management personnel to handle the abnormality in a timely manner, warnings can also be sent to the management personnel according to the summary analysis results, such as text messages, emails, etc., to ensure that the management personnel can receive and handle them in a timely manner.

[0039] When storing data, a classification storage strategy is adopted: all data in the data review process, including original abnormal data, external data, judgment results of the review model, and generated reports, etc., are classified and stored according to different categories. For example, the original data is stored in a data warehouse, and the review results and reports are stored in another independent database for subsequent query and management. According to the importance and usage frequency of the data, a reasonable data archiving strategy is formulated. For example, for the review data frequently used recently, it can be stored in a high-speed storage device for quick access; while for the earlier historical data, it can be archived to a lower-cost storage medium, such as a tape library or the cold storage area of cloud storage, while ensuring the security and integrity of the data. The archived data should be backed up and verified regularly to prevent data loss. To facilitate data query, an indexing system can be established. The index can be based on the key features of the data, such as employee identification code, abnormal data type, review time, etc., to improve the efficiency of data retrieval.

[0040] Correspondingly, this embodiment also provides a dynamic risk warning device for target personnel, such as Figure 2 shown, including: an edge computing device and a central server. The edge computing device is provided with: Data acquisition unit: used to acquire the behavior pattern data and external data of the target personnel; Model construction unit: create and train a behavior pattern recognition model based on the behavior pattern data, and create and train an abnormal data review model based on the behavior pattern data and external data; Abnormal data recognition unit: based on the behavior pattern recognition model, identify whether the current behavior pattern data of the target personnel is abnormal data; Abnormal data confirmation unit: based on the abnormal data and the external data corresponding to the abnormal data time period, use the abnormal data review model to confirm the abnormal data. If it is confirmed as abnormal data, upload the relevant information of the abnormal data to the central server.

[0041] The edge computing device also includes a working time determination unit, which is used to judge whether it is working time according to the task content of the target personnel and the behavior of the target personnel in the current time period; the data acquisition unit determines the data acquisition frequency according to whether it is working time.

[0042] The central server includes an information summary unit: used to summarize abnormal information; an alarm unit: used to warn the management personnel according to the summary analysis results; a storage unit, which stores data based on the classification storage strategy; an index unit, which is used to query the stored data.

Claims

1. A method for dynamically warning of the risks of target personnel, characterized in that, Including: Step 1: Deploy mobile edge computing devices; Step 2: Collect the behavior pattern data and external data of the target person based on the edge computing device, where the behavior pattern data includes the location data, behavior data and work-related data of the target person; the external data includes at least employee task data, customer visit history and expected duration, traffic data and meteorological data; Step 3: Create and train a behavior pattern recognition model based on the behavior pattern data, and create and train an abnormal data review model based on the behavior pattern data and external data; Step 4: Collect the behavior pattern data of the target person in real time, and use the behavior pattern recognition model for recognition. When it is detected that the current behavior pattern data is abnormal data, based on the abnormal data and the external data corresponding to the abnormal data time period, use the abnormal data review model to confirm the abnormal data. If it is confirmed as abnormal data, upload the relevant information of the abnormal data to the central server.

2. The dynamic risk warning method for target personnel according to claim 1, wherein The edge computing device includes a smart phone and a tablet computer.

3. The target person dynamic risk warning method according to claim 1, wherein Step 4 determines the data collection frequency according to whether the target person is at work.

4. The target person dynamic risk warning method according to claim 3, wherein Judge whether it is at work according to the task content of the target person and the behavior of the target person in the current time period.

5. The dynamic risk early warning method for target personnel according to claim 1, wherein It also includes Step 5: Summarize and analyze the relevant information of the abnormal data uploaded to the central server, and generate a risk distribution map and / or a risk report according to the summary analysis result.

6. The target person dynamic risk warning method according to claim 5, characterized in that Step 5 also includes: Warning the management personnel according to the summary analysis result.

7. Target personnel dynamic risk warning device, characterized in that Including: An edge computing device and a central server, and the following are provided on the edge computing device: A data collection unit: used to collect the behavior pattern data and external data of the target person; A model construction unit: Create and train a behavior pattern recognition model based on the behavior pattern data, and create and train an abnormal data review model based on the behavior pattern data and external data; An abnormal data recognition unit: Based on the behavior pattern recognition model, identify whether the current behavior pattern data of the target person is abnormal data; An abnormal data confirmation unit: Based on the abnormal data and the external data corresponding to the abnormal data time period, use the abnormal data review model to confirm the abnormal data. If it is confirmed as abnormal data, upload the relevant information of the abnormal data to the central server.

8. The target person dynamic risk warning device according to claim 7, characterized in that, The edge computing device also includes a working time determination unit, which is used to judge whether it is at work according to the task content of the target person and the behavior of the target person in the current time period; the data collection unit determines the data collection frequency according to whether it is at work.

9. The target person dynamic risk warning device according to claim 7 or 8, characterized in that, The central server includes an information summary unit: used to summarize abnormal information; an alarm unit: used to warn the management personnel according to the summary analysis result.

10. The target person dynamic risk warning device according to claim 9, characterized in that, The central server also includes a storage unit, which stores data based on a classification storage strategy; an index unit, which is used to query the stored data.

Images