Community work order intelligent management method and device

Abstract

The invention discloses a community work order intelligent management method and device, and relates to the technical field of community service management.The community work order intelligent management method comprises the steps that user work order data containing texts, voices and images are received through a multi-mode interface, and the user work order data are cached to a preprocessing queue; based on a semantic segmentation model, performing multi-dimensional analysis on the user work order data in the preprocessing queue, generating a structured work order comprising an equipment identifier, a fault feature and a geographic position, and associating a problem type label for the structured work order; according to the equipment identifier, the fault feature, the geographic position, the problem type label and an order division decision model, obtaining an assignment instruction of the structured order; and allocating the structured chemical worksheet to a corresponding processing terminal according to the allocation instruction, and updating the processing state of the structured chemical worksheet. According to the invention, the efficiency and response speed of community work order management are greatly improved.

Description

Technical Field

[0001] The present application relates to the field of community service management technology, and in particular to a method and device for intelligent management of community work orders. Background Art

[0002] With the acceleration of urbanization and the continuous expansion of communities, residents' demands for community services are becoming increasingly diverse and complex. Community managers need to quickly respond to residents' various issues, such as facility maintenance, environmental sanitation, and safety, to improve their life satisfaction. Therefore, an efficient and intelligent community work order management system has become a key requirement for community service management.

[0003] Currently, community work order management relies primarily on traditional manual processes. Residents submit issues via phone or on-site registration, community staff manually record and assign work orders, and progress is manually tracked and provided with feedback. This management approach can meet basic needs to a certain extent, but it is inefficient when faced with a large number of work orders.

[0004] Traditional work order management methods present numerous problems. First, manual recording and assigning of work orders is prone to errors and inefficient, making it difficult to respond to residents' needs in a timely manner. Second, the lack of effective data analysis and statistical functions makes it difficult to comprehensively monitor and optimize the processing of work orders. Furthermore, residents lack real-time information on the progress of work order processing, resulting in low satisfaction. Therefore, improving the efficiency and responsiveness of community work order management has become a pressing issue.

[0005] The above content is only used to assist in understanding the technical solution of this application and does not constitute an admission that the above content is prior art. Summary of the Invention

[0006] The purpose of this application is to provide a community work order intelligent management method and device, aiming to solve the technical problem of how to improve the efficiency and response speed of community work order management.

[0007] To achieve the above objectives, this application proposes a community work order intelligent management method, which includes:

[0008] Receiving user work order data including text, voice, and image through a multimodal interface, and caching the user work order data in a pre-processing queue;

[0009] Performing multi-dimensional analysis on the user work order data in the pre-processing queue based on a semantic segmentation model to generate a structured work order containing device identification, fault characteristics, and geographic location, and associating a problem type label with the structured work order;

[0010] Obtaining a dispatch instruction for the structured work order based on the equipment identification, the fault characteristics, the geographic location, the problem type label, and the order dispatch decision model;

[0011] The structured work order is allocated to a corresponding processing terminal according to the dispatch instruction, and the processing status of the structured work order is updated.

[0012] In one embodiment, the step of obtaining the dispatching instructions of the structured work order based on the equipment identification, the fault characteristics, the geographic location, the problem type label and the order splitting decision model includes: calculating the weights of the fault characteristics by an entropy weight method, and generating an emergency level classification result in combination with the population density parameter of the geographic location; matching the equipment skill library of the processing terminal based on the equipment identification, and screening a list of candidate terminals with corresponding equipment maintenance qualifications from the processing terminal; processing the problem type label, the emergency level classification result, the real-time load data of the candidate terminal list and the historical task completion rate based on the order splitting decision model to obtain the dispatching instructions of the structured work order including the processing terminal identification, task time limit and emergency alternative plan, and the order splitting decision model is constructed based on the Q-learning algorithm.

[0013] In one embodiment, the steps of calculating the weights of the fault features by the entropy weight method and generating the emergency level classification results in combination with the population density parameters of the geographical location include: calculating the entropy value of each fault feature based on the historical occurrence frequency and impact range of the fault features; assigning a weight to each fault feature based on the entropy value; dividing the regional emergency level according to the population density parameters of the geographical location and a preset population density threshold; and generating the emergency level classification result according to the weights of the fault features and the regional emergency level.

[0014] In one embodiment, the step of processing the problem type label, the emergency level classification result, the real-time load data of the candidate terminal list, and the historical task completion rate based on the order splitting decision model to obtain the dispatching instructions of the structured work order including the processing terminal identification, task time limit and emergency alternative plan includes: generating an initial work order processing priority based on the problem type label and the preset work order priority rule library; adjusting the initial work order processing priority by a nonlinear function according to the emergency level classification result to obtain a target work order processing priority; calculating the deviation of the real-time load data of each terminal in the candidate terminal list from the historical average load data, and generating a comprehensive processing capability score in combination with the sliding window mean of the historical task completion rate; matching the target work order processing priority and the comprehensive processing capability score through the order splitting decision model to generate the dispatching instructions of the structured work order including the processing terminal identification, task time limit and emergency alternative plan.

[0015] In one embodiment, the steps of constructing the order splitting decision model include: extracting the work order type, processing terminal identifier, actual task time and user satisfaction score from historical work order data to generate a multidimensional training feature set; constructing a state space based on the work order type and the processing terminal identifier, the state space including the work order urgency level, terminal real-time load and terminal skill matching degree; defining an action space based on the association between the processing terminal identifier and the work order type, the action space being a set of operations for assigning work orders to corresponding terminals; constructing a reward function based on the task time and the user satisfaction score, increasing the positive reward value of the reward function when the task time is lower than a preset time threshold, and superimposing the reward weight of the reward function when the user satisfaction score is higher than a preset score threshold; obtaining the target Q value table of the Q-learning algorithm based on the state space, the action space, the reward function and the candidate terminal list, and mapping the target Q value table to a fully connected neural network to obtain the order splitting decision model.

[0016] In one embodiment, the target Q value table of the Q-learning algorithm is obtained according to the state space, the action space, the reward function and the candidate terminal list, and the target Q value table is mapped to a fully connected neural network to obtain the order splitting decision model. The steps include: initializing the Q value table according to the dimension of the state space and the number of actions in the action space; determining the order splitting action according to the candidate terminal list and the Q value table using a greedy strategy; executing the order splitting action and calculating the reward value of the order splitting action according to the reward function; updating the corresponding Q value of the order splitting action in the Q value table according to a preset learning rate parameter, a preset discount factor parameter and the reward value; returning to the step of determining the order splitting action according to the candidate terminal list and the Q value table using a greedy strategy until the Q values of all state-action pairs in the Q value table meet the preset conditions to obtain the target Q value table; mapping the target Q value table to a fully connected neural network to obtain the order splitting decision model.

[0017] In one embodiment, the step of mapping the target Q value table to a fully connected neural network to obtain an order splitting decision model includes: extracting all state vectors and corresponding candidate terminal Q values from the target Q value table to construct a state-action training data set; constructing a fully connected neural network, the number of input layer nodes of the fully connected neural network is equal to the dimension of the state vector, and the number of output layer nodes of the fully connected neural network is equal to the number of candidate terminals in the candidate terminal list; using the state vector as the input feature and the candidate terminal Q value as the target label, training the fully connected neural network through supervised learning until the mean square error between the predicted Q value and the actual Q value is less than a preset error threshold, completing the training of the fully connected neural network; and using the trained fully connected neural network as the order splitting decision model.

[0018] In one embodiment, the steps of performing multi-dimensional analysis on the user work order data in the pre-processing queue based on the semantic segmentation model to generate a structured work order including equipment identification, fault characteristics and geographic location, and associating a problem type label with the structured work order include: performing keyword extraction on the text in the user work order data in the pre-processing queue to obtain equipment type description text; performing voice recognition on the voice in the user work order data to obtain fault description text; performing target detection on the image in the user work order data to obtain equipment identification and fault area image; inputting the equipment type description text, the fault description text and the fault area image into the semantic segmentation model for multi-dimensional analysis to generate a structured work order including the equipment identification, fault characteristics and geographic location; and associating a problem type label with the structured work order based on the matching result between the fault description text and a preset problem type library.

[0019] In one embodiment, after the step of assigning the structured work order to the corresponding processing terminal according to the dispatch instruction and updating the processing status of the structured work order, it also includes: when receiving user feedback data, extracting the work order number in the user feedback data; retrieving the work order attachment and processing node operation log corresponding to the work order number from the work order attachment library and the operation log library; performing correlation analysis on the user feedback data, the work order attachment and the processing node operation log to obtain feedback analysis results and fault feature types; and adjusting the historical occurrence frequency weight and impact range weight of the corresponding fault feature in the entropy weight method according to the feedback analysis results and the fault feature type.

[0020] In addition, to achieve the above objectives, the present application also proposes a community work order intelligent management device, which includes:

[0021] A data receiving module is used to receive user work order data including text, voice and image through a multimodal interface, and cache the user work order data in a pre-processing queue;

[0022] A semantic parsing module, configured to perform multi-dimensional parsing of the user work order data in the pre-processing queue based on a semantic segmentation model, generate a structured work order containing device identification, fault characteristics, and geographic location, and associate a problem type label with the structured work order;

[0023] An intelligent dispatching module, configured to obtain dispatch instructions for the structured work order based on the equipment identification, the fault characteristics, the geographical location, and the order dispatching decision model;

[0024] The work order execution module is used to allocate the structured work order to the corresponding processing terminal according to the dispatch instruction and update the processing status of the structured work order.

[0025] In addition, to achieve the above-mentioned purpose, the present application also proposes a community work order intelligent management device, which includes: a memory, a processor, and a computer program stored on the memory and runnable on the processor, and the computer program is configured to implement the steps of the community work order intelligent management method as described above.

[0026] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium. A computer program is stored on the storage medium, and when the computer program is executed by the processor, the steps of the community work order intelligent management method as described above are implemented.

[0027] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the steps of the community work order intelligent management method as described above.

[0028] One or more technical solutions proposed in this application have at least the following technical effects:

[0029] First, the community intelligent AI work order service platform receives work order data submitted by users, including text, voice, and images, through a multimodal interface and caches this data in a preprocessing queue. This allows users to submit work orders in a variety of ways, improving user experience and submission efficiency while avoiding data loss. Next, the platform performs multi-dimensional analysis of the work order data in the preprocessing queue based on a semantic segmentation model, generating structured work orders that include device identification, fault characteristics, and geographic location, and associating problem type labels with the work orders. This process not only improves the accuracy and standardization of work order information, but also provides a foundation for subsequent intelligent order sorting through automatic classification, greatly improving the efficiency and responsiveness of community work order management. The platform then generates dispatch instructions for the structured work order based on the device identification, fault characteristics, geographic location, problem type labels, and the order sorting decision model. Intelligent dispatching ensures that work orders are assigned to the most appropriate processing terminal or processing personnel based on actual conditions, avoiding the subjectivity and inefficiency of manual dispatching and further improving the professionalism and accuracy of processing. Finally, the platform assigns the structured work order to the corresponding processing terminal according to the dispatch instructions and updates the processing status of the work order. This step ensures that maintenance personnel can obtain work order information in a timely manner and start processing. At the same time, it allows managers and users to understand the processing progress of the work order in real time, improves transparency and user experience, and facilitates subsequent statistical analysis and feedback collection. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0031] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0032] Figure 1 A flowchart of the first embodiment of the community work order intelligent management method provided in this application;

[0033] Figure 2 Schematic diagram of the community intelligent AI work order service platform provided for Example 1 of the community work order intelligent management method of this application;

[0034] Figure 3 A schematic diagram of work order submission provided in Example 1 of the community work order intelligent management method of this application;

[0035] Figure 4 A work order management diagram provided for Example 1 of the community work order intelligent management method of this application;

[0036] Figure 5 A schematic diagram of the intelligent order splitting results provided in Example 1 of the community work order intelligent management method of this application;

[0037] Figure 6 A flowchart of the second embodiment of the community work order intelligent management method provided in this application;

[0038] Figure 7 A user feedback data management diagram provided for Example 2 of the community work order intelligent management method of this application;

[0039] Figure 8 This is a schematic diagram of the module structure of the community work order intelligent management device according to the embodiment of the present application;

[0040] Figure 9 This is a schematic diagram of the device structure of the hardware operating environment involved in the community work order intelligent management method in the embodiment of this application.

[0041] The purpose, features and advantages of this application will be further explained with reference to the accompanying drawings in conjunction with the embodiments. DETAILED DESCRIPTION

[0042] It should be understood that the specific embodiments described herein are merely used to explain the technical solutions of the present application and are not intended to limit the present application.

[0043] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0044] With the acceleration of urbanization, the scale of communities and the growing demand for resident services are driving the development of efficient and intelligent community work order management systems, which are crucial for improving resident satisfaction. Currently, community work order management relies on manual processes, such as submitting issues by phone or on-site, manually recording assignments, and tracking feedback. This approach is inefficient and prone to errors when handling large numbers of work orders. This traditional approach is not only slow to respond to resident needs, lacks data analysis and statistical capabilities to optimize management, but also prevents residents from real-time monitoring of work order progress, leading to low resident satisfaction.

[0045] The main solution of the embodiment of the present application is: receiving text, voice and image work order data submitted by users through a multimodal interface, and caching it in a pre-processing queue to improve submission efficiency and user experience. Use a semantic segmentation model to parse this data and generate a structured work order containing equipment identification, fault characteristics, geographic location and problem type labels. Based on information such as equipment identification, fault characteristics and the order distribution decision model, the platform intelligently distributes work orders to the most suitable processing terminal or personnel. Finally, the work order processing status is updated according to the distribution instructions to ensure that the work order is processed in a timely and effective manner.

[0046] It should be noted that the execution entity of the embodiments of this application can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, mobile phone, etc., or an electronic device capable of implementing the above functions, a community intelligent AI work order service platform, etc. The following uses the community intelligent AI work order service platform as an example to illustrate this embodiment and the following embodiments.

[0047] Based on this, the embodiment of the present application provides a community work order intelligent management method, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the community work order intelligent management method of this application.

[0048] In this embodiment, the community work order intelligent management method includes steps S10 to S40:

[0049] Step S10: receiving user work order data including text, voice, and image through a multimodal interface, and caching the user work order data into a pre-processing queue.

[0050] Please note that, please refer to Figure 2 , Figure 2 This diagram shows the community intelligent AI work order service platform provided in Example 1 of the community work order intelligent management method of this application. The diagram illustrates the platform's user interface and key functional modules. The top section displays a data overview, including device statistics, content statistics, and data volume statistics. Key indicators include the total number of devices, daily work order submissions, daily work order pick-ups, completed work orders, and user satisfaction ratings. The middle section includes data statistics and detailed analysis, such as the daily number of work orders processed, the number of pending work orders, user satisfaction ratings, average response time, and the number of daily activated users. This data helps platform administrators monitor work order processing efficiency and user satisfaction. The data analysis module below displays the daily number of work orders processed and completion rate, while the data statistics table lists detailed information such as the work order number, submission time, work order status, priority, and handling personnel. In addition, the platform provides early warning information and the latest announcements to ensure timely response to potential issues and maintain communication between users and administrators. Overall, this diagram demonstrates the platform's comprehensive capabilities in improving community service response speed, optimizing work order management, and enhancing user experience.

[0051] A multimodal interface is an interactive interface or system interface that can accept multiple types of input data. In the Community Intelligent AI Work Order Service Platform, multimodal interfaces allow users to submit work order information in a variety of ways, including text input (such as filling out a work order description), voice input (such as leaving a voice message or voice command), and image upload (such as taking a photo of the problem scene).

[0052] User work order data refers to all the information provided by users when submitting work orders, including but not limited to problem description, urgency, relevant pictures, voice recordings, etc. This data is a key source of information for community managers to understand residents' needs and problems.

[0053] Please refer to Figure 3 , Figure 3 The work order submission diagram provided for the first embodiment of the community work order intelligent management method of this application shows the work order submission process interface, through which users can intelligently generate work orders, fill in information, upload attachments, and set priorities. The figure shows a pop-up window with four main steps: First, "Intelligent Work Order Generation" uses AI technology to automatically create work orders, reducing the need for manual filling; second, "Work Order Information Filling" allows users to enter detailed work order content, such as request type and description; then, "Work Order Attachment Upload" supports users to add relevant files or pictures to provide a more comprehensive description of the problem; finally, "Work Order Priority Setting" allows users to set the priority of the work order according to the urgency (if the user does not enter, the platform automatically generates it based on the work order information) to ensure that key issues are handled in a timely manner. In addition, the to-do statistics chart on the left side of the interface shows the number of work orders of different priorities, helping users quickly understand the distribution of work orders. The entire process is designed to improve the efficiency and accuracy of work order submission, thereby optimizing the response speed and quality of community services.

[0054] The pre-processing queue is a buffer area that temporarily stores user work order data. It is used to perform preliminary processing and sorting of data before the work order is officially processed. This ensures that the data can be quickly received and temporarily stored before entering the formal processing flow, avoiding data loss or processing delays.

[0055] It's understandable that, first, the Community Intelligent AI Ticket Service Platform receives user-submitted ticket data through its multimodal interface. This data may include text descriptions, voice messages, and image files. Receiving this data ensures that the platform can comprehensively collect user problem information and accurately receive it regardless of the user's chosen input method. Secondly, the platform sends the received user ticket data to a preprocessing queue for caching. This serves to temporarily store the data before formal processing, avoiding data loss or processing delays while preparing for subsequent data processing and analysis.

[0056] Step S20: Perform multi-dimensional analysis on the user work order data in the pre-processing queue based on the semantic segmentation model to generate a structured work order including device identification, fault characteristics, and geographic location, and associate a problem type label with the structured work order.

[0057] It should be noted that the semantic segmentation model is a deep learning algorithm used to classify and analyze image or text content at the pixel level or semantic level.

[0058] Device identification refers to the unique identification information of the specific device mentioned in the user work order. This can be the device name, model, number or other characteristics that can clearly distinguish the device.

[0059] Fault characteristics refer to the specific manifestations or problem characteristics of the equipment failure described in the user work order. This can be specific descriptions such as the equipment not being able to work properly, abnormal sounds, or error codes being displayed.

[0060] Geographic location refers to the physical location of the device or issue mentioned in a user's work order. This can be a specific address, floor, room number, or device installation location. For example, if a user mentions "water meter failure in Room 102, Unit 3, Building 1" in a work order, "Room 102, Unit 3, Building 1" is the geographic location.

[0061] Structured work orders refer to work order information that has been parsed and organized and stored in a standardized format. It contains key information such as equipment identification, fault characteristics, and geographic location. This information is stored in the database in a clear and standardized format.

[0062] Problem type tags are categorized automatically or manually based on the work order content to identify the specific type of problem. For example, if the work order description is "Unusual noise during elevator operation," the system will automatically associate the problem type tag "Elevator failure" to assign the work order to elevator maintenance personnel.

[0063] It is understandable that first, the community intelligent AI work order service platform reads user-submitted work order data from the preprocessing queue. Next, the platform uses a semantic segmentation model to deeply analyze this data, extracting key information such as device identification, fault characteristics, and geographic location. This information is organized and stored in a structured work order format to ensure the efficiency and accuracy of subsequent processing. Finally, the platform associates the corresponding problem type label with the structured work order based on the extracted key information, allowing the system to quickly identify the work order category and automatically sort and prioritize it, thereby improving the efficiency and accuracy of work order processing.

[0064] Please refer to Figure 4 , Figure 4The work order management diagram provided for the first embodiment of the community work order intelligent management method of this application lists in detail the key information of the work order, such as the serial number, work order ID, declarant, work order status, priority, and creation time. Users can manage work orders through this interface, including viewing work order details, transferring work orders, editing work order information, and downloading related attachments. The interface displays work orders of different statuses, such as "Processing", "Completed" and "New", as well as work orders of different priorities, such as "High", "Medium" and "Low", allowing users to quickly identify and process urgent and important work orders. In addition, the interface also provides the functions of adding, importing, exporting and deleting work orders, which is convenient for users to perform batch operations and management of work orders. Through this interface, platform administrators can effectively monitor the processing progress of work orders, optimize resource allocation, improve the response speed and processing efficiency of community services, and ensure that the needs of community residents are met in a timely manner.

[0065] As an example, the steps of performing multi-dimensional analysis on the user work order data in the pre-processing queue based on the semantic segmentation model to generate a structured work order containing equipment identification, fault characteristics and geographic location, and associating a problem type label with the structured work order include: performing keyword extraction on the text in the user work order data in the pre-processing queue to obtain equipment type description text; performing voice recognition on the voice in the user work order data to obtain fault description text; performing target detection on the image in the user work order data to obtain equipment identification and fault area image; inputting the equipment type description text, the fault description text and the fault area image into the semantic segmentation model for multi-dimensional analysis to generate a structured work order containing the equipment identification, fault characteristics and geographic location; and associating a problem type label with the structured work order based on the matching result between the fault description text and a preset problem type library.

[0066] Device type description text refers to descriptive information about the device type extracted from the text portion of the user's work order data. For example, if a user mentions in a work order that "the smart access control system at the community gate is malfunctioning," "smart access control system" is the device type description text.

[0067] Fault description text refers to a detailed description of the equipment fault, extracted from the spoken portion of user work order data using voice recognition technology. For example, if a user leaves a voice message saying "The elevator makes an unusual noise," the fault description text obtained through voice recognition is "The elevator makes an unusual noise."

[0068] Device identification refers to the unique identification information of a device, identified from the image portion of a user's work order data using object detection technology. For example, if a user uploads an image with a device number or model, the object detection model can identify the device number or model in the image, such as "water meter number 12345." Device identification clearly identifies the specific device involved in the work order, ensuring that maintenance personnel can quickly locate the device that needs attention.

[0069] A fault area image is an image of the equipment fault area identified by object detection technology from the image portion of the user's work order data. For example, if a user uploads a photo showing a faulty part of the equipment, the object detection model can identify the fault area and crop it to create a fault area image. This image helps maintenance personnel intuitively understand the specific location and condition of the fault.

[0070] The preset problem type library is a collection of common problem types and their descriptions, predefined and stored by the system. For example, the library might include labels such as "elevator failure," "access control system failure," and "water meter failure," along with their corresponding common descriptions. When the system parses the problem description text in a user's work order, it matches it with the content in the preset problem type library, associating the most suitable problem type label with the structured work order. This step helps automatically categorize and quickly dispatch work orders, improving processing efficiency.

[0071] First, the community intelligent AI work order service platform extracts user-submitted text from a pre-processing queue. Using natural language processing (NLP) keyword extraction algorithms, it accurately locates vocabulary and phrases related to device types and generates a device type description to clearly identify the specific device involved in the work order. Secondly, the platform performs voice recognition on the user-submitted voice information, converting the voice signal into text to generate a detailed fault description. This converts the user's verbal description into readable text, facilitating subsequent processing and analysis. The platform then performs object detection on the user-submitted image information, identifying key elements in the image, such as the device identifier and fault area, and crops the fault area to create a fault area image. This allows the platform to intuitively capture the device's appearance and fault location. The device type description, fault description, and fault area image are then fed into a semantic segmentation model. The model then analyzes this information and, combined with other data such as geolocation, generates a structured work order containing the device identifier, fault characteristics, and geolocation. This structured work order makes the work order information more standardized and easier to process. Finally, the platform compares the fault description text with a preset problem type library and associates the corresponding problem type label with the structured work order based on the matching results. This process realizes the automatic classification of work orders, improves the efficiency and accuracy of work order processing, and enables subsequent maintenance personnel to quickly understand the nature of the work order and take appropriate measures.

[0072] Step S30: Obtaining a dispatch instruction for the structured work order based on the equipment identification, the fault characteristics, the geographical location, the problem type label, and the order dispatch decision model.

[0073] It should be noted that the order allocation decision model refers to the model or algorithm used in the community intelligent AI work order service platform to determine how to allocate work orders. It determines the processing terminal, processing person, or processing team to which the work order should be assigned based on multiple key information of the work order (such as device identification, fault characteristics, geographic location, and problem type label). The order allocation decision model typically considers factors such as the processing terminal or processing person's professional skills, current workload, and geographic location to ensure that work orders are efficiently and accurately allocated to the most suitable processing terminal or processing person.

[0074] The dispatch instruction refers to the specific allocation command generated by the work order decision model based on the key information of the work order. It clearly specifies which processing terminal or processing personnel the work order should be assigned to, as well as the relevant processing priority and processing requirements.

[0075] It is understandable that, first, the community intelligent AI work order service platform extracts key information from structured work orders, including equipment identification, fault characteristics, geographic location, and problem type labels. Secondly, the platform inputs this extracted information into the order splitting decision model. The model will conduct a comprehensive evaluation and matching of work orders based on preset rules and algorithms, combined with factors such as the professional skills, current workload, and geographic location of the processing terminal or processing personnel, to ensure that the work order can be assigned to the most suitable processing terminal or processing personnel. Finally, the order splitting decision model generates a dispatch instruction based on the evaluation results, clearly specifies the assignment of work orders to specific processing terminals or processing personnel, and determines the priority of processing, thereby ensuring that work orders can enter the processing process efficiently and accurately, improving response speed and processing efficiency.

[0076] Step S40: Allocate the structured work order to a corresponding processing terminal according to the dispatch instruction, and update the processing status of the structured work order.

[0077] It should be noted that processing terminals refer to the specific devices or systems used in the community intelligent AI work order service platform to receive, process, and feedback work order information. These terminals typically include maintenance personnel's mobile devices (such as mobile phones and tablets), workstation computers, and possibly smart devices (such as smart access control systems and elevator control systems). Processing terminals are the actual execution point for work order processing. Maintenance personnel use these terminals to receive work order information, perform on-site processing, and feedback the processing results to the system.

[0078] The processing status refers to the current progress and status of a structured work order during its processing. It reflects the various stages of a work order from generation to completion, such as "To be Assigned," "Assigned," "To Be Processed," "Processing," and "Completed." Updating the processing status is a crucial component of a work order management system, ensuring that managers and users have real-time visibility into the progress of work orders, improving transparency and user experience. For example, when a maintenance technician begins processing a work order, the system updates the status to "Processing." Once the repair is complete and feedback is received, the status is updated to "Completed."

[0079] It is understandable that, first of all, the community intelligent AI work order service platform accurately locates the corresponding processing terminal based on the dispatch instructions, ensuring that the work order information is accurately transmitted to the maintenance personnel's mobile device or workbench computer. The purpose of this is to allow maintenance personnel to obtain the work order details at the first time and respond quickly. Secondly, the platform synchronously updates the work order status from "pending assignment" to "assigned". This update operation is to allow managers and users to grasp the dynamics of the work order in real time and avoid information lag. Finally, after the maintenance personnel completes the task, they submit the processing results on the processing terminal, and the platform immediately updates the work order status to "completed". This step is to ensure that the entire work order processing process has a beginning and an end, which is convenient for subsequent statistical analysis and user feedback collection.

[0080] Please refer to Figure 5 , Figure 5 This is a schematic diagram of the intelligent order splitting results provided for Example 1 of the community work order intelligent management method of this application, which lists the detailed information of multiple work orders, including key data such as work order number, applicant, problem type, priority, etc. Each work order is clearly marked with its problem type, such as network failure, software problem, hardware failure, etc., and the corresponding priority, from high to low, to help platform administrators quickly identify and handle urgent problems. Users can view the details of each work order by clicking the "Details" button, modify the work order information through the "Edit" button, or remove no longer needed work orders through the "Delete" button. In addition, the interface also provides the function of adding new work orders, which is convenient for users to submit new service requests according to actual conditions. The design of this interface is to improve the transparency and efficiency of work order processing, ensure that the problems of community residents can be solved in a timely and effective manner, thereby improving the overall service quality and resident satisfaction.

[0081] This embodiment provides a method for intelligent management of community work orders. First, the community intelligent AI work order service platform receives work order data containing text, voice, and images submitted by users through a multimodal interface, and caches this data in a pre-processing queue. This allows users to submit work orders in a variety of ways, improving user experience and submission efficiency while avoiding data loss. Next, the platform performs multi-dimensional analysis of the work order data in the pre-processing queue based on a semantic segmentation model, generates a structured work order containing device identification, fault characteristics, and geographic location, and associates a problem type label with the work order. This process not only improves the accuracy and standardization of work order information, but also provides a basis for subsequent intelligent order sorting through automatic classification, greatly improving the efficiency and response speed of community work order management. Then, the platform obtains the dispatch instructions for the structured work order based on the device identification, fault characteristics, geographic location, problem type label, and order sorting decision model. Intelligent dispatching ensures that the work order can be assigned to the most suitable processing terminal or processing personnel according to the actual situation, avoiding the subjectivity and inefficiency of manual dispatching, and further improving the professionalism and accuracy of the processing. Finally, the platform assigns the structured work order to the corresponding processing terminal according to the dispatch instructions and updates the processing status of the work order. This step ensures that maintenance personnel can obtain work order information in a timely manner and start processing. At the same time, it allows managers and users to understand the processing progress of the work order in real time, improves transparency and user experience, and facilitates subsequent statistical analysis and feedback collection.

[0082] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above embodiment 1 can be referred to the above introduction and will not be described in detail later. Figure 6 , Figure 6 This is a flow chart of the second embodiment of the community work order intelligent management method of this application. Step S30 of the community work order intelligent management method includes steps S31 to S33:

[0083] Step S31 , weighting the fault features is calculated using an entropy weight method, and an emergency level classification result is generated in combination with the population density parameter of the geographical location.

[0084] It should be noted that the entropy weight method is a multi-attribute decision analysis method based on information entropy, which is used to determine the weight of each attribute (or feature). The smaller the information entropy, the greater the degree of variation of the feature, and the higher its importance in decision making.

[0085] The population density parameter is the ratio of the population within a specific geographic location to the area of that location, reflecting the density of people in that location. For example, a high population density in a certain area within a community (such as the residential gate or elevator lobby) means that the processing of work orders in that area may affect the lives of more residents.

[0086] The urgency classification results are based on a comprehensive assessment of the fault feature weights and the population density of the location. These levels are typically categorized into multiple levels (such as "High," "Medium," and "Low") to indicate the priority level for handling the work order. For example, a work order located in a densely populated area with a high fault feature weight would be classified as "High" and thus prioritized.

[0087] It is understandable that, first, the community intelligent AI work order service platform collects fault feature data from user work orders, uses the entropy weight method to calculate the weights of these fault features, and determines the weight of each feature by analyzing the degree of variation of the fault features. The greater the degree of variation, the higher the weight, thus more accurately reflecting the importance of each fault feature in the work order urgency assessment. Secondly, the platform obtains the geographic location information of the fault location and queries the population density parameters of the location. Areas with high population density mean that the work order has a wider impact and a greater impact on residents' lives, so they need to be given higher priority in the urgency level assessment. Finally, the platform combines the calculated fault feature weights with the population density parameters of the geographic location. According to preset rules and models, it comprehensively evaluates the urgency of the work order, generates an emergency level classification result, and divides the work order into different emergency levels to rationally allocate resources, prioritize the processing of high-urgency work orders, and improve the efficiency and response speed of community work order management.

[0088] As an example, the steps of calculating the weights of the fault features by the entropy weight method and generating the emergency level classification results in combination with the population density parameters of the geographical location include: calculating the entropy value of each fault feature based on the historical occurrence frequency and impact range of the fault features; assigning a weight to each fault feature based on the entropy value; dividing the regional emergency level according to the population density parameters of the geographical location and the preset population density threshold; and generating the emergency level classification result according to the weights of the fault features and the regional emergency level.

[0089] The historical occurrence frequency refers to the ratio of the number of times a certain fault feature occurs in the past period of time to the total number of work orders. It reflects the commonness of the fault feature and provides data support for evaluating its importance.

[0090] Impact scope refers to the degree to which a fault characteristic affects the lives of community residents. This can be assessed based on factors such as the type of equipment involved, frequency of use, and the number of residents affected. For example, an elevator failure may affect the mobility of residents throughout a building, while a water meter failure may only affect a few households. The larger the impact scope, the more severe the fault characteristic's impact on community residents' lives, and the more it warrants priority treatment.

[0091] Entropy is a quantitative representation of information entropy, which is used to measure the uncertainty or variation of fault characteristics. By calculating the information entropy of each fault characteristic, its entropy value can be obtained. The calculation formula of information entropy is:

[0092]

[0093] Among them, H refers to information entropy, p i represents the probability of fault feature i appearing in all work orders, and n is the number of fault features.

[0094] The entropy value E is the normalized form of information entropy, and the calculation formula is:

[0095]

[0096] Where H refers to information entropy and n is the number of fault features. Based on the entropy value, the weight of each fault feature can be calculated:

[0097]

[0098] Among them, E i is the entropy value of the i-th fault feature, E j is the entropy value of the j-th fault feature, and the weight w i It reflects the importance of fault feature i.

[0099] The preset population density threshold refers to one or more population density reference values set based on the actual situation of the community, which is used to divide the emergency levels of different areas.

[0100] Regional emergency level refers to the emergency level of different areas based on the population density parameters of the geographical location and the preset population density threshold.

[0101] First, the Community Intelligent AI Work Order Service Platform extracts the frequency and impact of each fault feature from historical work order data. By analyzing this data, it calculates the entropy value for each fault feature. Features with lower entropy values are more important in work order processing. Second, the platform assigns a weight to each fault feature based on the calculated entropy value. The weight directly reflects the fault feature's importance in assessing urgency; higher weights indicate a greater impact on the work order's urgency. The platform then classifies different areas into urgency levels, combining geographic population density parameters with pre-set population density thresholds. High-density areas typically have higher urgency levels because work orders in these areas are likely to affect more residents. Finally, the platform comprehensively considers the fault feature weights and regional urgency levels to generate a final urgency classification, classifying work orders into different urgency levels. This allows for more efficient resource allocation and prioritizes work orders with higher urgency, thereby improving the efficiency and response speed of community work order management.

[0102] As an example, after the step of assigning the structured work order to the corresponding processing terminal according to the dispatch instruction and updating the processing status of the structured work order, it also includes: when receiving user feedback data, extracting the work order number in the user feedback data; retrieving the work order attachment and processing node operation log corresponding to the work order number from the work order attachment library and the operation log library; performing correlation analysis on the user feedback data, the work order attachment and the processing node operation log to obtain feedback analysis results and fault feature types; and adjusting the historical occurrence frequency weight and impact range weight of the corresponding fault feature in the entropy weight method based on the feedback analysis results and the fault feature type.

[0103] User feedback data refers to the feedback information submitted by users after the work order processing is completed, including satisfaction evaluation of the work order processing results, evaluation of maintenance personnel, suggestions or opinions on the processing process, etc.

[0104] The work order number is a unique identifier for each work order in the system, used to distinguish different work orders. It is usually a combination of numbers or letters, which makes it easier for the system and users to quickly find and manage work orders.

[0105] The ticket attachment library is a database or storage system that stores all attachments related to a ticket. These attachments may include user-submitted images, videos, documents, etc., which are used to support the ticket description or provide more detailed information.

[0106] An operation log library is a database or storage system that records all operations during the ticket processing process. These operation logs include detailed information such as the ticket's creation time, assignment time, processing time, processing personnel, and processing steps. Operation logs are used to track the ticket processing process and facilitate subsequent auditing and analysis.

[0107] Work order attachments refer to auxiliary files uploaded by users or processing personnel during the work order processing process, such as photos, videos, maintenance reports, etc.

[0108] The processing node operation log refers to the operation record of the work order at different processing stages, including the timestamp, operator, operation content, etc. of each processing node.

[0109] Feedback analysis results refer to the conclusions drawn from analyzing user feedback data, including user satisfaction scores, a summary of common issues, and improvement suggestions. These results are used to evaluate the quality of ticket processing and user satisfaction, providing a basis for platform optimization.

[0110] Fault signature types are categorized based on user feedback and work order processing records. For example, fault signature types might be "device aging," "human damage," or "software failure." These categorizations help the platform better understand the nature of faults and optimize handling strategies.

[0111] The historical frequency weight is calculated based on the historical frequency of the fault characteristic in the entropy weighting method. This weight reflects how frequently the fault characteristic has appeared in past work orders. The higher the frequency, the greater the weight. The impact range weight is calculated based on the impact range of the fault characteristic in the entropy weighting method. This weight reflects the degree of impact of the fault characteristic on the lives of community residents. The larger the impact range, the greater the weight.

[0112] Please refer to Figure 7 , Figure 7 A user feedback data management diagram is provided for Example 2 of the community work order intelligent management method of this application, which lists the feedback information submitted by users, including key data such as feedback ID, feedback user, and feedback time. User feedback is an important basis for the platform to optimize services and enhance user experience. Through these feedbacks, the platform can understand the satisfaction with work order processing, existing problems, and specific suggestions from users. Each feedback entry provides edit and delete options to facilitate administrators to manage and maintain feedback information. In addition, the interface also provides the function of adding new data, allowing administrators to add new user feedback records. By effectively managing and utilizing these feedback data, the platform can continuously improve the work order processing process, improve service quality, and enhance user satisfaction, thereby achieving more intelligent and efficient community service management.

[0113] First, when the community intelligent AI work order service platform receives user feedback data, the system will automatically parse the data and extract the work order number to ensure that it can be accurately linked to the specific work order record. Secondly, the system uses the extracted work order number to retrieve the work order attachment corresponding to the number from the work order attachment library; at the same time, it retrieves the processing node operation log of the work order from the operation log library. These logs record in detail every step of the operation and time node in the work order processing process. Then, the system will correlate and analyze the user feedback data, work order attachments, and processing node operation logs. By comparing the problem description in the user feedback with the actual fault manifestation in the work order attachment, combined with the processing flow in the operation log, the specific problem points and fault feature types of the user feedback are analyzed. For example, if the user feedback is "incomplete repair", the system will combine the fault photos in the attachment and the repair steps in the operation log to determine which link has the problem. Finally, based on the feedback analysis results and fault feature types obtained through correlation analysis, the system will adjust the historical frequency weight and impact range weight of the corresponding fault feature in the entropy weight method. If the analysis results show that a certain fault feature frequently appears in user feedback and has a wide impact range, then the system will increase the weight of the fault feature so that more attention can be paid to it in subsequent work order processing; conversely, if a certain fault feature has a low frequency of occurrence and a small impact range, the system will reduce its weight, thereby optimizing the emergency level assessment and processing strategy of subsequent work orders, and improving the efficiency and response speed of community work order management.

[0114] Step S32 : matching the equipment skill library of the processing terminal based on the equipment identification, and screening a list of candidate terminals with corresponding equipment maintenance qualifications from the processing terminal.

[0115] It should be noted that the equipment skill library refers to a database stored in the community intelligent AI work order service platform, which records the equipment maintenance skills and qualifications of each processing terminal. This information includes the maintenance personnel's professional skills, certification qualifications, and the types of equipment and fault ranges they can handle.

[0116] Equipment maintenance qualifications refer to the certification that maintenance personnel or processing terminals possess the legal and effective ability to repair specific equipment. These qualifications are typically issued by relevant industry organizations and demonstrate that maintenance personnel are capable of resolving specific equipment failures.

[0117] The candidate terminal list refers to a list of processing terminals with corresponding equipment maintenance qualifications selected based on the equipment skill library in the community intelligent AI work order service platform.

[0118] It's understandable that, first, after receiving a work order, the community intelligent AI work order service platform extracts the device identification information from the work order as the key basis for querying. The platform then uses the device identification to perform a matching query within the device skill library, screening out processing terminals with the corresponding equipment maintenance qualifications. Finally, the platform compiles the screened processing terminals into a candidate terminal list, providing a precise selection range for subsequent work order assignments. This ensures that work orders are efficiently and accurately assigned to maintenance personnel with the corresponding capabilities, thereby improving the professionalism and responsiveness of work order processing.

[0119] Step S33: Based on the order splitting decision model, the problem type label, the emergency level classification result, the real-time load data of the candidate terminal list, and the historical task completion rate are processed to obtain the dispatch instructions of the structured order including the processing terminal identification, task deadline and emergency alternative plan. The order splitting decision model is constructed based on the Q-learning algorithm.

[0120] It should be noted that real-time load data refers to the workload of the processing terminal at the current point in time, including information such as the number of work orders being processed and the estimated completion time. The historical task completion rate refers to the proportion of tasks completed by the processing terminal over a period of time, reflecting the processing terminal's efficiency and reliability. The processing terminal ID is a unique identification code for each processing terminal, used to distinguish different processing terminals. It is usually a number or device ID, which facilitates the platform to specify the recipient of work orders when assigning them.

[0121] The deadline is the time within which a work order must be completed. This deadline is set based on the urgency and difficulty of the work order, ensuring that maintenance personnel can complete the task within a reasonable time.

[0122] Emergency backup plans are backup solutions that are prepared in case the primary processing terminal fails to complete a task on time or experiences a malfunction. For example, if a maintenance technician is unable to arrive on site on time, the system can automatically assign the work order to a nearby maintenance technician or provide remote technical support.

[0123] The Q-learning algorithm is a reinforcement learning algorithm used to train intelligent agents (such as order splitting decision models) to make optimal decisions in complex environments.

[0124] As you can understand, the platform first collects real-time load data from processing terminals to understand the current workload of each terminal. Then, the platform uses a Q-learning-based order-dispatching decision model, combining problem type labels, urgency level classification results, and historical task completion rates, to perform comprehensive processing. Based on this data, the model generates dispatch instructions that include the processing terminal identifier, task deadline, and emergency backup plan. The processing terminal identifier ensures that work orders are accurately assigned to the appropriate maintenance personnel, the task deadline sets a time limit for maintenance personnel to complete the task, and the emergency backup plan provides a backup solution for possible emergencies, thereby ensuring that work orders are completed efficiently and accurately.

[0125] As an example, the step of processing the problem type label, the emergency level classification result, the real-time load data of the candidate terminal list, and the historical task completion rate based on the order splitting decision model to obtain the dispatching instructions of the structured work order including the processing terminal identification, task deadline and emergency alternative plan includes: generating an initial work order processing priority based on the problem type label and the preset work order priority rule library; adjusting the initial work order processing priority through a nonlinear function according to the emergency level classification result to obtain a target work order processing priority; calculating the deviation of the real-time load data of each terminal in the candidate terminal list from the historical average load data, and generating a comprehensive processing capability score in combination with the sliding window mean of the historical task completion rate; matching the target work order processing priority and the comprehensive processing capability score through the order splitting decision model to generate the dispatching instructions of the structured work order including the processing terminal identification, task deadline and emergency alternative plan.

[0126] The default work order priority rule base is a database that stores various problem type labels and corresponding initial priority rules. These rules predefine the initial priority of work orders based on the nature of the work order (such as equipment type and fault severity), providing a basis for subsequent priority adjustments.

[0127] The initial work order processing priority is the preliminary priority of the work order generated based on the preset work order priority rule library. It reflects the processing priority of the work order without considering other dynamic factors (such as urgency level, real-time load, etc.).

[0128] The nonlinear function is a mathematical function used to adjust the initial work order processing priority based on the urgency level classification results. The formula is as follows:

[0129] P target =P initial ×(1+α·E 2 )

[0130] Among them, P initialRefers to the initial work order processing priority, E refers to the emergency level classification result, P target is the adjusted target work order processing priority, and α is an adjustment coefficient used to control the impact of the urgency level on the priority.

[0131] The target ticket processing priority is the final ticket processing priority after being adjusted by a nonlinear function. It comprehensively considers the issue type and urgency level, accurately reflecting the actual processing priority of the ticket.

[0132] Historical average load data refers to the average workload of the processing terminal over the past period of time. It is used to evaluate the long-term workload of the processing terminal and provide a reference for comparison with real-time load data.

[0133] Deviation refers to the difference between the real-time load data of the processing terminal and the historical average load data. It reflects the degree of deviation between the current load and the long-term average load and is used to assess the current working pressure of the processing terminal.

[0134] The sliding window mean refers to the average task completion rate of the processing terminal in the recent period, which can reflect the long-term performance of the processing terminal.

[0135] The comprehensive processing capability score is generated based on the deviation and sliding window mean of the processing terminal. It is used to evaluate the comprehensive processing capability of the processing terminal. This score takes into account the current load and historical performance of the processing terminal, providing a more comprehensive reference for the order splitting decision model.

[0136] First, the community intelligent AI work order service platform searches for the corresponding priority rule in the preset work order priority rule library based on the work order's problem type label, and directly assigns the work order an initial processing priority value. For example, for the problem type of "elevator failure", the rule library may preset its initial priority to 8 (out of 10). This is done to preliminarily determine the urgency of the work order based on the nature of the problem. Secondly, based on the work order's urgency level classification results, the platform adjusts the initial priority through a preset nonlinear function. Assuming the urgency level is "high", the corresponding nonlinear function will increase the initial priority to 9. This can more accurately reflect the actual urgency of the work order and ensure that urgent work orders receive higher priority processing. The platform then calculates the deviation between the real-time load data and the historical average load data for each terminal in the candidate terminal list. If a terminal's current load is significantly higher than its historical average, the deviation is large. Combined with the sliding window average of the terminal's historical task completion rate, the platform comprehensively evaluates the terminal's current workload and long-term performance, generating a comprehensive processing capability score. For example, a terminal with a low deviation and a high sliding window average indicates low current workload and excellent long-term performance, resulting in a high comprehensive processing capability score. This provides a comprehensive measure of the terminal's processing capability. Finally, through the order allocation decision model, the target work order processing priority is matched with the comprehensive processing capability score of each candidate terminal. The terminal with the highest score is selected as the work order processing target, and a dispatch instruction is generated, including the processing terminal identifier, task deadline, and emergency backup plan. The task deadline is set based on the urgency of the work order, and the emergency backup plan provides a backup solution for possible emergencies. The entire process ensures that work orders can be efficiently and accurately assigned to the most suitable processing terminal, improving the efficiency and response speed of work order processing.

[0137] As an example, the steps of constructing the order splitting decision model include: extracting the work order type, processing terminal identifier, actual task time and user satisfaction score from historical work order data to generate a multidimensional training feature set; constructing a state space based on the work order type and the processing terminal identifier, the state space including the work order urgency level, terminal real-time load and terminal skill matching degree; defining an action space based on the association between the processing terminal identifier and the work order type, the action space being a set of operations for assigning work orders to corresponding terminals; constructing a reward function based on the task time and the user satisfaction score, increasing the positive reward value of the reward function when the task time is lower than a preset time threshold, and superimposing the reward weight of the reward function when the user satisfaction score is higher than a preset score threshold; obtaining the target Q value table of the Q-learning algorithm based on the state space, the action space, the reward function and the candidate terminal list, and mapping the target Q value table to a fully connected neural network to obtain the order splitting decision model.

[0138] Historical work order data refers to the detailed records of previous work orders saved in the system.

[0139] The work order type refers to the classification of the problem or task involved in the work order, such as "elevator failure", "access control system failure", "water pipe repair", etc.

[0140] The processing terminal identifier refers to a unique identification code for each processing terminal, which is used to distinguish different processing terminals.

[0141] The actual time consumed by a task refers to the actual time required for a processing terminal to complete a work order. It reflects the efficiency of work order processing and is an important indicator for evaluating the performance of the processing terminal.

[0142] User satisfaction score refers to the user's evaluation of the work order processing results, usually a value (such as 1 to 10) or a level (such as "very satisfied", "satisfied", "unsatisfied").

[0143] The multi-dimensional training feature set is a collection of features extracted from historical work order data and used to train the order splitting decision model. These features include work order type, processing terminal identifier, actual task duration, and user satisfaction rating, providing a rich data foundation for the model.

[0144] The state space refers to the set of all possible states in the order splitting decision model. In the community intelligent AI work order service platform, the state space includes the work order urgency level, the terminal's real-time load, and the terminal's skill matching. These states reflect the current status of work order processing and the processing terminal's capabilities.

[0145] The urgency level of a work order is a rating based on the urgency of the work order, typically categorized as "high," "medium," or "low." Work orders with a high urgency level are prioritized to minimize impact on users.

[0146] Real-time terminal load refers to the workload of the processing terminal at the current point in time, including the number of work orders being processed and the estimated completion time. This data is used to assess the current workload of the processing terminal and ensure the rationality of work order allocation.

[0147] Terminal skill match refers to the degree to which the skills required to process a terminal match those of a work order. For example, a terminal with elevator maintenance qualifications has a high skill match for an elevator fault work order. Terminals with high skill match are more likely to complete work orders efficiently.

[0148] An association relationship refers to the correspondence between a processing terminal identifier and a work order type, that is, which processing terminals have the ability to process a specific work order type. This relationship is used to determine which terminals can process a specific type of work order.

[0149] The action space refers to the set of all possible actions in the order splitting decision model. In the community intelligent AI work order service platform, the action space is the set of operations for assigning work orders to corresponding terminals.

[0150] An operation set refers to the set of all possible operations for assigning a work order to a corresponding terminal. For example, the system can choose to assign a work order to terminal A or terminal B, etc. These choices constitute an operation set.

[0151] A reward function is a function used in reinforcement learning to evaluate the quality of actions. It evaluates the rationality of ticket assignments based on task duration and user satisfaction scores. Assignments with shorter task durations and higher user satisfaction receive higher rewards.

[0152] The preset time threshold refers to the maximum task completion time pre-set by the system. If the actual task time is lower than this threshold, it means that the work order processing efficiency is high, and the reward function will increase the positive reward value.

[0153] A positive reward value refers to the additional reward value given in the reward function when the task duration is lower than the preset time threshold. This value is used to motivate the system to select efficient allocation actions.

[0154] The preset scoring threshold refers to the minimum value of the user satisfaction score preset by the system. If the user satisfaction score is higher than this threshold, it means that the service quality is high, and the reward function will add reward weight.

[0155] Reward weight refers to the additional weight given in the reward function when the user satisfaction score is higher than the preset score threshold. This weight is used to further incentivize the platform to select allocation actions with high user satisfaction.

[0156] The target Q-value table is a Q-value table trained using the Q-learning algorithm. It contains the Q-values corresponding to each state and action. The Q-value reflects the expected reward for selecting an action in a given state. The target Q-value table is the core component of the order splitting decision model, used to guide the allocation of work orders.

[0157] A fully connected neural network is a neural network structure in which each neuron is connected to all neurons in the previous layer. It is used to approximate the target Q-value table and map the state space and action space to Q-values, thereby achieving efficient order splitting decisions.

[0158] First, the platform extracts work order type, processing terminal ID, actual task duration, and user satisfaction rating from historical work order data. This data is integrated into a multidimensional training feature set, providing a data foundation for subsequent model training. Second, the platform constructs a state space based on the work order type and processing terminal ID. Specifically, the platform uses the work order urgency level, terminal real-time load, and terminal skill match as dimensions of the state space, with each dimension corresponding to a value or state. For example, the work order urgency level can be categorized as high, medium, and low. The terminal real-time load can be represented by the current number of tasks, and the terminal skill match is calculated based on the terminal's skill library and the work order type. This state space is constructed to comprehensively describe the current state of work order processing and the capabilities of the processing terminal. Then, the platform defines an action space based on the association between the processing terminal ID and the work order type. The action space is the set of all possible actions that can be taken to assign a work order to a corresponding terminal. Specifically, all processing terminals are listed. For each terminal, if its skill match meets the requirements of the work order type, it is added to the action space. This definition of the action space clarifies the actions that the model can choose in each state. Next, the platform constructs a reward function based on task duration and user satisfaction scores. When the task duration is below the preset time threshold, the positive reward value of the reward function is increased. For example, if the time threshold is set to 2 hours and the task duration is less than 2 hours, the reward value is increased by 10 points. When the user satisfaction score is higher than the preset score threshold, the reward weight of the reward function is superimposed. For example, if the score threshold is set to 8 points and the user satisfaction score is higher than 8 points, the reward weight is increased by 0.5. This reward function is constructed to incentivize the model to select efficient and high-quality work order assignment actions. Finally, based on the state space, action space, reward function, and candidate terminal list, the platform uses the Q-learning algorithm to train the target Q-value table, and maps the Q-value table to a fully connected neural network to train the order assignment decision model. The purpose of this is to use the powerful fitting ability of the neural network to map the state space and action space to Q values, thereby achieving intelligent and efficient work order assignment decisions.

[0159] As an example, the target Q value table of the Q-learning algorithm is obtained according to the state space, the action space, the reward function and the candidate terminal list, and the target Q value table is mapped to a fully connected neural network to obtain the order splitting decision model. The steps include: initializing the Q value table according to the dimension of the state space and the number of actions in the action space; determining the order splitting action according to the candidate terminal list and the Q value table using a greedy strategy; executing the order splitting action and calculating the reward value of the order splitting action according to the reward function; updating the corresponding Q value of the order splitting action in the Q value table according to a preset learning rate parameter, a preset discount factor parameter and the reward value; returning to the step of determining the order splitting action according to the candidate terminal list and the Q value table using a greedy strategy until the Q values of all state-action pairs in the Q value table meet the preset conditions to obtain the target Q value table; mapping the target Q value table to a fully connected neural network to obtain the order splitting decision model.

[0160] A greedy strategy is one that always chooses the action with the highest Q value in the current state when selecting an action. This strategy aims to maximize immediate rewards, such as the ε-greedy strategy.

[0161] The order splitting action refers to the specific operation of assigning a work order to a certain processing terminal. In the Q-learning algorithm, the order splitting action is a specific action selected from the action space, such as assigning the work order to terminal A or terminal B.

[0162] The reward value is a numerical value calculated based on the reward function and is used to evaluate the quality of a particular order splitting action. The higher the reward value, the greater the immediate benefit brought by the action.

[0163] The preset learning rate parameter is a key parameter in the Q-learning algorithm, controlling how quickly new information updates old information. The learning rate determines the step size of the Q-value update. A higher learning rate results in faster Q-value updates, but can lead to unstable learning processes. A lower learning rate results in slower Q-value updates, but a more stable learning process.

[0164] The preset discount factor parameter is another important parameter in the Q-learning algorithm. It controls the degree to which future rewards are discounted. The discount factor determines the impact of future rewards on the current Q value. The closer the discount factor is to 1, the greater the impact of future rewards on the current Q value; the closer the discount factor is to 0, the smaller the impact of future rewards on the current Q value.

[0165] The Q value represents the expected reward of selecting a certain action in a certain state. The higher the Q value, the greater the expected benefit of selecting the action in that state. The Q value gradually approaches the optimal value through continuous learning and updating. The Q value update formula is:

[0166]

[0167] Where s represents the current state, which is a vector describing the current work order processing status. a represents the action selected in the current state s. The action is a specific operation selected from the action space. α is the preset learning rate parameter, γ is the preset discount factor parameter, R is the reward value of the current action, s′ is the new state after executing the action, and a′ represents the optimal action among all possible actions in the new state s′.

[0168] A state-action pair is a combination of selecting a specific action in a specific state. Each state-action pair has a corresponding Q value, which represents the expected reward of selecting the action in that state.

[0169] The precondition is the convergence condition for the Q-learning algorithm, typically set to a Q-value change of less than a preset convergence threshold for N consecutive iterations. For example, if N = 10 and the convergence threshold is 0.01, the algorithm is considered to have converged when the Q-value change is less than 0.01 for 10 consecutive iterations, and training stops.

[0170] First, the platform initializes the Q-value table based on the dimensions of the state space and the number of actions in the action space, setting the Q-value of each state-action pair to 0. Next, the platform uses the ε-greedy strategy to determine the order splitting action. Specifically, it generates a random number between 0 and 1. If the random number is less than or equal to 0.1, it randomly selects a processing terminal identifier from the candidate terminal list as the current action to explore new states and actions. If the random number is greater than 0.1, it queries the Q-values of all actions corresponding to the state space in the current Q-table and selects the processing terminal identifier corresponding to the action with the largest Q-value as the current action to leverage the learned knowledge. The platform then executes the selected order splitting action and calculates the reward value for the action based on a reward function that evaluates the quality of the action based on task duration and user satisfaction scores. The platform then updates the Q-value corresponding to the order splitting action in the Q-value table based on the preset learning rate parameter (e.g., 0.1), the discount factor parameter (e.g., 0.9), and the calculated reward value. The update formula is:

[0171] Q(s,a) = Q(s,a) + learning rate * (reward value + discount factor * maxQ(s',a') - Q(s,a))

[0172] Where s is the current state, a is the current action, s' is the new state after executing the action, and a' represents all possible actions in the new state. Finally, the platform repeats this process until the Q-value change for all state-action pairs in the Q-value table is less than 0.01 for 10 consecutive iterations. The Q-value table is considered to have converged, and the target Q-value table is obtained. Finally, the platform maps the target Q-value table to a fully connected neural network. Using the neural network's learning capabilities, it further optimizes the order splitting decision, resulting in the final order splitting decision model, which is used to guide actual work order allocation.

[0173] As an example, the step of mapping the target Q value table to a fully connected neural network to obtain an order splitting decision model includes: extracting all state vectors and corresponding candidate terminal Q values from the target Q value table to construct a state-action training data set; constructing a fully connected neural network, the number of input layer nodes of the fully connected neural network is equal to the dimension of the state vector, and the number of output layer nodes of the fully connected neural network is equal to the number of candidate terminals in the candidate terminal list; using the state vector as the input feature and the candidate terminal Q value as the target label, training the fully connected neural network through supervised learning until the mean square error between the predicted Q value and the actual Q value is less than a preset error threshold, completing the training of the fully connected neural network; and using the trained fully connected neural network as the order splitting decision model.

[0174] The state vector is a set of feature values that describe the current status of work order processing, including the work order's urgency level, the terminal's real-time load, and the terminal's skill match. These feature values are combined to form a vector that represents the current state. The candidate terminal Q-value refers to the Q-value of the action corresponding to each candidate terminal in the target Q-value table. These Q-values represent the expected reward for selecting a candidate terminal as the processing terminal under the current state. A state-action training dataset, consisting of state vectors and the corresponding candidate terminal Q-values, is used to train fully connected neural networks, enabling them to learn the relationship between state and action.

[0175] Input features refer to the input data of the fully connected neural network, namely the state vector. The target label refers to the output target of the fully connected neural network, namely the Q-value of the candidate terminal. It represents the expected reward for selecting a candidate terminal as the processing terminal in the current state and is used to guide neural network training. Predicted Q-values refer to the Q-values predicted by the fully connected neural network based on the input state vector during training. These predicted values are compared with the actual Q-values to evaluate the effectiveness of model training. The actual Q-values are Q-values extracted from the target Q-value table. These values are obtained through training with the Q-learning algorithm and represent the expected reward for selecting a candidate terminal as the processing terminal in the current state. The preset error threshold is an upper limit on the error set during training of the fully connected neural network. When the mean squared error between the predicted and actual Q-values is less than this threshold, the model training is considered complete and has achieved the expected accuracy.

[0176] First, the platform extracts all state vectors and their corresponding candidate terminal Q-values from the target Q-value table and organizes these data into a state-action training dataset. The state vector, consisting of the work order's urgency level, the terminal's real-time load, and the terminal's skill match, describes the specific circumstances of each work order, while the candidate terminal Q-value represents the expected reward assigned to each terminal under that state. This dataset organization allows the model to learn the relationship between states and actions. Secondly, the platform constructs a fully connected neural network. The number of input layer nodes matches the dimensionality of the state vector to ensure the network receives complete state information. The number of output layer nodes matches the number of terminals in the candidate terminal list, enabling the network to output a predicted Q-value for each candidate terminal. The network's intermediate layers are configured with multiple hidden layers and nodes as needed to enhance the network's learning capabilities. Then, the platform trains the fully connected neural network through supervised learning, using the state vector as input features and the candidate terminal Q-values as target labels. The training process uses mean squared error as the loss function. The network weights are continuously adjusted through a backpropagation algorithm until the mean squared error between the predicted and actual Q-values falls below a preset error threshold. This completes the neural network training process, enabling the network to accurately predict the Q-values of each action under a given state. Finally, the trained fully connected neural network is used as the order allocation decision model. In actual applications, the model receives the real-time state vector as input and outputs the Q-value prediction of each candidate terminal. The work order is allocated by selecting the processing terminal corresponding to the maximum Q-value, thereby realizing intelligent and efficient work order allocation decision-making.

[0177] This embodiment first calculates the weights of the fault features of the work order using the entropy weight method, determines the importance weights based on the historical frequency of the fault features and the scope of impact, and then generates the urgency level classification results of the work order in combination with the population density parameter of the geographical location. This allows the urgency of the work order to be accurately assessed based on the severity of the fault and the number of people affected, ensuring that faults in high-impact areas are handled first. Secondly, based on the device identification in the work order, a list of candidate terminals with corresponding equipment maintenance qualifications is matched and screened from the equipment skill library of the processing terminal. This ensures that only maintenance personnel with the corresponding skills and qualifications are considered for work order assignment, thereby ensuring the professionalism and reliability of the maintenance work. Finally, a work order dispatching decision model based on the Q-learning algorithm is used to comprehensively consider the problem type label, urgency level classification results, real-time load data of the candidate terminal list, and historical task completion rate to generate structured work order dispatching instructions that include processing terminal identification, task deadlines, and emergency alternative plans. This not only allows tasks to be reasonably allocated based on the urgency of the work order and the actual workload of maintenance personnel, but also provides contingency plans for possible emergencies through emergency alternative plans, further improving the efficiency of work order dispatching and the ability to deal with complex situations, thereby realizing intelligent and efficient work order management for the entire community.

[0178] This application also provides a community work order intelligent management device, please refer to Figure 8 , the community work order intelligent management device includes:

[0179] The data receiving module 10 is configured to receive user work order data including text, voice, and image through a multimodal interface, and cache the user work order data in a pre-processing queue;

[0180] A semantic parsing module 20 is configured to perform multi-dimensional parsing of the user work order data in the pre-processing queue based on a semantic segmentation model, generate a structured work order including device identification, fault characteristics, and geographic location, and associate a problem type label with the structured work order;

[0181] An intelligent dispatching module 30 is configured to obtain dispatch instructions for the structured work order based on the equipment identification, the fault characteristics, the geographical location, and the order dispatching decision model;

[0182] The work order execution module 40 is configured to allocate the structured work order to a corresponding processing terminal according to the dispatch instruction and update the processing status of the structured work order.

[0183] The community work order intelligent management device provided in this application adopts the community work order intelligent management method in the above-mentioned embodiment, which can solve the technical problem of how to improve the efficiency and response speed of community work order management. Compared with the existing technology, the beneficial effects of the community work order intelligent management device provided in this application are the same as the beneficial effects of the community work order intelligent management method provided in the above-mentioned embodiment, and the other technical features of the community work order intelligent management device are the same as the features disclosed in the above-mentioned embodiment method, which will not be repeated here.

[0184] The present application provides a community work order intelligent management device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the community work order intelligent management method in the above-mentioned embodiment one.

[0185] Reference below Figure 9 , which shows a schematic diagram of the structure of a community work order intelligent management device suitable for implementing the embodiments of the present application. The community work order intelligent management device in the embodiments of the present application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 9 The community work order intelligent management device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0186] like Figure 9As shown, the community work order intelligent management device may include a processing device 1001 (such as a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to the program stored in the ROM (Read Only Memory) 1002 or the program loaded from the storage device 1003 to the RAM (Random Access Memory) 1004. Various programs and data required for the operation of the community work order intelligent management device are also stored in RAM1004. The processing device 1001, ROM1002 and RAM1004 are connected to each other via a bus 1005. The input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 1008 including, for example, an LCD (Liquid Crystal Display), a speaker, a vibrator, etc.; storage devices 1003 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 1009. The communication device 1009 can allow the community work order intelligent management device to communicate wirelessly or wired with other devices to exchange data. Although the figure shows a community work order intelligent management device with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems may be implemented or have instead.

[0187] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program comprising program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0188] The community work order intelligent management device provided by this application adopts the community work order intelligent management method in the above embodiment, which can solve the technical problem of how to improve the efficiency and response speed of community work order management. Compared with the existing technology, the beneficial effects of the community work order intelligent management device provided by this application are the same as the beneficial effects of the community work order intelligent management method provided by the above embodiment, and the other technical features of the community work order intelligent management device are the same as the features disclosed in the previous embodiment method, which will not be repeated here.

[0189] The present application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the community work order intelligent management method described in the above-described embodiment. The computer-readable storage medium may be included in a community work order intelligent management device, or may exist independently without being incorporated into the community work order intelligent management device.

[0190] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by the community work order intelligent management device, the community work order intelligent management device enables the following: to receive user work order data containing text, voice and images through a multimodal interface, and cache the user work order data to a preprocessing queue; to perform multi-dimensional analysis on the user work order data in the preprocessing queue based on a semantic segmentation model, to generate a structured work order containing equipment identification, fault characteristics and geographic location, and to associate a problem type label with the structured work order; to obtain a dispatch instruction for the structured work order based on the equipment identification, the fault characteristics, the geographic location, the problem type label and the order dispatch decision model; to assign the structured work order to the corresponding processing terminal according to the dispatch instruction, and to update the processing status of the structured work order.

[0191] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a module, program segment or a part of code, and the module, program segment or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented by a dedicated hardware-based system that performs the specified function or operation, or can be implemented by a combination of dedicated hardware and computer instructions.

[0192] The readable storage medium provided in this application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., a computer program) for executing the above-mentioned community work order intelligent management method, and can solve the technical problem of how to improve the efficiency and response speed of community work order management. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the community work order intelligent management method provided in the above-mentioned embodiment, and will not be repeated here.

[0193] The present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the above-mentioned community work order intelligent management method.

[0194] The computer program product provided in this application can solve the technical problem of how to improve the efficiency and response speed of community work order management. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the community work order intelligent management method provided above, and will not be repeated here.

Claims

1. A community work order intelligent management method, characterized in that: The method comprises: Receiving user work order data including text, voice, and image through a multimodal interface, and caching the user work order data in a pre-processing queue; Performing multi-dimensional analysis on the user work order data in the pre-processing queue based on a semantic segmentation model to generate a structured work order containing device identification, fault characteristics, and geographic location, and associating a problem type label with the structured work order; Obtaining a dispatch instruction for the structured work order based on the equipment identification, the fault characteristics, the geographic location, the problem type label, and the order dispatch decision model; The structured work order is allocated to a corresponding processing terminal according to the dispatch instruction, and the processing status of the structured work order is updated.

2. The method according to claim 1, wherein The step of obtaining the dispatch instruction of the structured work order according to the equipment identification, the fault characteristics, the geographical location, the problem type label, and the order dispatch decision model includes: The fault characteristics are weighted by an entropy weight method, and an emergency level classification result is generated by combining the population density parameter of the geographical location; Based on the equipment identification, the equipment skill library of the processing terminal is matched, and a list of candidate terminals with corresponding equipment maintenance qualifications is screened from the processing terminal; Based on the order splitting decision model, the problem type label, the emergency level classification result, the real-time load data of the candidate terminal list and the historical task completion rate are processed to obtain the dispatch instructions of the structured order including the processing terminal identification, task deadline and emergency alternative plan. The order splitting decision model is constructed based on the Q-learning algorithm.

3. The method according to claim 2, wherein The step of calculating the weights of the fault features by the entropy weight method and generating an emergency level classification result in combination with the population density parameter of the geographical location includes: Calculating the entropy value of each fault feature according to the historical occurrence frequency and impact range of the fault feature; assigning a weight to each of the fault features based on the entropy value; Classify the regional emergency level according to the population density parameter of the geographical location and the preset population density threshold; The emergency level classification result is generated according to the weight of the fault feature and the regional emergency level.

4. The method according to claim 2, wherein The step of processing the problem type label, the emergency level classification result, the real-time load data of the candidate terminal list, and the historical task completion rate based on the order splitting decision model to obtain the dispatch instruction of the structured order including the processing terminal identifier, the task time limit, and the emergency alternative plan includes: Generate an initial work order processing priority based on the problem type label and a preset work order priority rule library; According to the emergency level classification result, the initial work order processing priority is adjusted by a nonlinear function to obtain a target work order processing priority; Calculating the deviation between the real-time load data and the historical average load data of each terminal in the candidate terminal list, and combining it with the sliding window mean of the historical task completion rate to generate a comprehensive processing capability score; The target work order processing priority and the comprehensive processing capability score are matched through the order splitting decision model to generate a dispatch instruction for the structured work order including the processing terminal identification, task time limit and emergency alternative plan.

5. The method according to claim 2, wherein The steps of constructing the order splitting decision model include: Extract the work order type, processing terminal identifier, actual task duration, and user satisfaction score from historical work order data to generate a multi-dimensional training feature set; Constructing a state space based on the work order type and the processing terminal identifier, the state space including the work order urgency level, the terminal real-time load, and the terminal skill matching degree; Defining an action space according to the association between the processing terminal identifier and the work order type, wherein the action space is a set of operations for assigning the work order to the corresponding terminal; Constructing a reward function based on the task duration and the user satisfaction score, increasing the positive reward value of the reward function when the task duration is lower than a preset time threshold, and adding a reward weight of the reward function when the user satisfaction score is higher than a preset score threshold; A target Q value table of the Q-learning algorithm is obtained according to the state space, the action space, the reward function and the candidate terminal list, and the target Q value table is mapped to a fully connected neural network to obtain an order splitting decision model.

6. The method according to claim 5, wherein The step of obtaining a target Q-value table of a Q-learning algorithm based on the state space, the action space, the reward function, and the candidate terminal list, and mapping the target Q-value table to a fully connected neural network to obtain an order splitting decision model includes: Initializing a Q-value table according to the dimension of the state space and the number of actions in the action space; Adopting a greedy strategy to determine the order splitting action according to the candidate terminal list and the Q value table; Executing the order splitting action and calculating the reward value of the order splitting action according to the reward function; Update the Q value corresponding to the order splitting action in the Q value table according to the preset learning rate parameter, the preset discount factor parameter and the reward value; Return to the step of using a greedy strategy to determine the order splitting action according to the candidate terminal list and the Q value table, until the Q values of all state-action pairs in the Q value table meet the preset conditions, and obtain the target Q value table; The target Q value table is mapped to a fully connected neural network to obtain an order splitting decision model.

7. The method according to claim 6, wherein The step of mapping the target Q value table to a fully connected neural network to obtain an order splitting decision model includes: Extract all state vectors and corresponding candidate terminal Q values from the target Q value table to construct a state-action training dataset; Constructing a fully connected neural network, wherein the number of input layer nodes of the fully connected neural network is equal to the dimension of the state vector, and the number of output layer nodes of the fully connected neural network is equal to the number of candidate terminals in the candidate terminal list; Using the state vector as an input feature and the candidate terminal Q value as a target label, the fully connected neural network is trained through supervised learning until the mean square error between the predicted Q value and the actual Q value is less than a preset error threshold, thereby completing the training of the fully connected neural network; The trained fully connected neural network is used as the order splitting decision model.

8. The method according to claim 1, wherein The steps of performing multi-dimensional analysis on the user work order data in the pre-processing queue based on the semantic segmentation model to generate a structured work order including the device identification, fault characteristics, and geographic location, and associating the structured work order with a problem type label include: Perform keyword extraction on the text in the user work order data in the pre-processing queue to obtain a device type description text; Performing voice recognition on the voice in the user work order data to obtain a fault description text; Performing target detection on the image in the user work order data to obtain a device identification and a fault area image; Input the device type description text, the fault description text, and the fault area image into a semantic segmentation model for multi-dimensional analysis to generate a structured work order containing the device identification, fault characteristics, and geographic location; According to the matching result of the fault description text and the preset problem type library, a problem type label is associated with the structured work order.

9. The method according to any one of claims 2 to 7, characterized in that After the steps of allocating the structured work order to the corresponding processing terminal according to the dispatch instruction and updating the processing status of the structured work order, the method further includes: When receiving user feedback data, extracting the work order number in the user feedback data; Retrieving the work order attachment and processing node operation log corresponding to the work order number from the work order attachment library and operation log library; Performing correlation analysis on the user feedback data, the work order attachments, and the processing node operation logs to obtain feedback analysis results and fault feature types; According to the feedback analysis result and the fault feature type, the historical occurrence frequency weight and the impact range weight of the corresponding fault feature in the entropy weight method are adjusted.

10. A community work order intelligent management device, characterized in that: The device comprises: A data receiving module is used to receive user work order data including text, voice and image through a multimodal interface, and cache the user work order data in a pre-processing queue; A semantic parsing module, configured to perform multi-dimensional parsing of the user work order data in the pre-processing queue based on a semantic segmentation model, generate a structured work order containing device identification, fault characteristics, and geographic location, and associate a problem type label with the structured work order; An intelligent dispatching module, configured to obtain dispatch instructions for the structured work order based on the equipment identification, the fault characteristics, the geographical location, and the order dispatching decision model; The work order execution module is used to allocate the structured work order to the corresponding processing terminal according to the dispatch instruction and update the processing status of the structured work order.

Images