Work order distribution method, device, equipment, medium and product

Abstract

The embodiment of the invention provides a work order distribution method and device, equipment, a medium and a product, and relates to the technical field of computers. The work order allocation method comprises the following steps: acquiring a work order feature vector of a target work order and personnel feature vectors of a plurality of candidate processing personnel corresponding to the target work order; inputting the work order feature vector of the target work order and the personnel of the plurality of candidate processing personnel into a matching model, and matching the target work order with the plurality of candidate processing personnel to obtain the matching degree of the target work order and each candidate processing personnel; wherein the matching model takes a work order feature vector of a sample work order corresponding to the target work order and a personnel feature vector of a candidate processing personnel corresponding to the sample work order as input, and a matching degree label of the sample work order and the candidate processing personnel corresponding to the sample work order is obtained by pre-training an output label; and based on the matching degree between the target work order and each candidate processing person, allocating the processing person to the target work order.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a work order allocation method, apparatus, equipment, medium and product. Background Technology

[0002] Pre-sales work orders for ICT (Information and Communications Technology) solutions, such as those for IoT (Internet of Things) or communication technology solutions, are highly complex and ambiguous, typically requiring specialized technical personnel for processing. Currently, these work orders are usually assigned manually to designated experts based on pre-defined rules. This method is inefficient and heavily reliant on the experience of the assigning personnel. Inexperienced personnel or poor matching between assigned and task-specific experts can negatively impact the quality and efficiency of work order processing. Summary of the Invention

[0003] This application provides a work order allocation method to improve the allocation efficiency and adaptability of personnel handling work orders such as pre-sales work orders for ICT technology solutions.

[0004] In a first aspect, embodiments of this application provide a work order allocation method, the method comprising: Obtain the work order feature vector of the target work order and the personnel feature vectors of multiple candidate handlers corresponding to the target work order; wherein, the target work order is the work order for which handlers are to be assigned; the work order feature vector of the target work order is matched with the personnel feature vectors of the multiple candidate handlers using a matching model, and the matching degree between the target work order and each candidate handler is obtained; wherein, the matching model is pre-trained using the work order feature vector of the sample work order corresponding to the target work order and the personnel feature vector of the candidate handler corresponding to the sample work order as input, and the matching degree label between the sample work order and the candidate handler corresponding to the sample work order as output label; based on the matching degree between the target work order and each candidate handler, handlers are assigned to the target work order.

[0005] In some embodiments of this application, the matching model includes: a multi-head attention model and a multilayer perceptron model; the step of matching the target work order's work order feature vector with the personnel input matching model of the multiple candidate processing personnel to match the target work order with the multiple candidate processing personnel, and obtaining the matching degree between the target work order and each of the candidate processing personnel, includes: The work order feature vector of the target work order and the personnel feature vectors of the multiple candidate processing personnel are input into a multi-head attention model for multi-head attention calculation to obtain a fusion vector corresponding to each candidate processing personnel; the fusion vector corresponding to each candidate processing personnel is input into a multilayer perceptron model for matching degree calculation to obtain a matching degree corresponding to each candidate processing personnel.

[0006] In some embodiments of this application, the matching model is updated periodically, and the method further includes: The accuracy rate of personnel allocation in the previous model cycle is calculated. If the allocation accuracy rate is less than the accuracy rate threshold, the service quality information of historical work orders in the previous model cycle is obtained. The service quality information includes the work order problem resolution status and user satisfaction. Based on the service quality information of the historical work orders, the matching degree between the personnel handling the historical work orders and the historical work orders is determined. Using the work order feature vector and the personnel feature vector of the personnel handling the historical work orders as inputs, and the matching degree between the personnel handling the historical work orders and the historical work orders as the output label, the matching model of the previous model cycle is updated to obtain the matching model of the current model cycle.

[0007] In some embodiments of this application, the step of obtaining the work order feature vector of the target work order and the personnel corresponding to the target work order and multiple candidate processing personnel includes: The feature values ​​of multiple features of the target work order are obtained. Based on the feature value of each feature, the work order feature vector library corresponding to the target work order is queried. Based on the feature vector corresponding to the feature value of each feature, the work order feature vector is generated. Based on the personnel identifier of each candidate processing personnel, the personnel feature vector library corresponding to the target work order is queried to obtain the personnel feature vector of each candidate processing personnel.

[0008] In some embodiments of this application, before obtaining the personnel feature vectors of multiple candidate processing personnel corresponding to the target work order, the method further includes: Based on the domain of the user's problem in the target work order, multiple candidate handlers corresponding to the target work order type are screened to obtain multiple first candidate handlers; wherein, the target work order type is the work order type to which the target work order belongs; the first candidate handlers are those whose professional fields match the domain of the user's problem among the multiple candidate handlers corresponding to the target work order type; based on the current work order load of the first candidate handlers and their corresponding personal work order load threshold, the multiple first candidate handlers are screened to obtain multiple second candidate handlers; the second candidate handlers are those whose current work order load is less than their corresponding personal work order load threshold among the multiple first candidate handlers; based on the multiple second candidate handlers, multiple candidate handlers corresponding to the target work order are determined.

[0009] In some embodiments of this application, the individual work order load threshold is updated periodically, and the method further includes: Based on the efficiency information and service quality information of the candidate processing personnel in the previous threshold period, the efficiency information and business fluctuation information of the candidate processing personnel in this threshold period, and the level information of the candidate processing personnel, the individual work order load threshold of the candidate processing personnel in this threshold period is determined.

[0010] Secondly, embodiments of this application provide a work order allocation device, the device comprising: The module includes an acquisition module for acquiring the work order feature vector of the target work order and the personnel feature vectors of multiple candidate processing personnel corresponding to the target work order; a matching module for matching the work order feature vector of the target work order with the personnel input of the multiple candidate processing personnel into a matching model, and matching the target work order with the multiple candidate processing personnel to obtain the matching degree between the target work order and each candidate processing personnel; wherein, the matching model is pre-trained using the work order feature vector of the sample work order corresponding to the target work order and the personnel feature vector of the candidate processing personnel corresponding to the sample work order as input, and the matching degree label between the sample work order and the candidate processing personnel corresponding to the sample work order as the output label; and an allocation module for allocating processing personnel to the target work order based on the matching degree between the target work order and each candidate processing personnel.

[0011] Thirdly, embodiments of this application provide a work order allocation device, the device comprising: A memory that stores a computer program; a processor for executing the computer program to implement the steps of the above-described work order allocation method.

[0012] Fourthly, embodiments of this application provide a readable storage medium storing a computer program that can be executed by a processor to implement the steps of the above-described work order allocation method.

[0013] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described work order allocation method.

[0014] The technical solution provided in this application, by obtaining the work order feature vector of a target work order (such as a pre-sales work order for a technical solution) and the personnel feature vectors of multiple candidate processing personnel, matches the target work order with the multiple candidate processing personnel using a matching model trained on the work order feature vector of the target work order and the multiple candidate processing personnel, thereby obtaining the matching degree between the target work order and each of the candidate processing personnel; and assigning processing personnel to the target work order based on the matching degree between the target work order and each of the candidate processing personnel. This enables automatic allocation of processing personnel for pre-sales work orders of technical solutions, improving the efficiency and accuracy of personnel allocation. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating the work order allocation method of some embodiments of this application; Figure 2 This is a schematic diagram of the work order allocation device according to some embodiments of this application. Detailed Implementation

[0017] This application provides a work order allocation method, apparatus, equipment, product, and medium.

[0018] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this application.

[0019] Figure 1This is a flowchart illustrating the work order allocation method of some embodiments of this application. For example... Figure 1 As shown, in some embodiments of this application, the work order allocation method includes the following steps: S102, obtain the work order feature vector of the target work order and the personnel feature vectors of multiple candidate processing personnel corresponding to the target work order.

[0020] The target work order is the work order to be assigned to a specific person for processing. This work order can be a pre-sales work order for ICT technology solutions. The work order feature vector is used to characterize the target work order. It is a multi-dimensional vector, where each dimension represents a feature of the target work order.

[0021] In practice, feature values ​​of multiple features of the target work order can be obtained, each feature value can be vectorized to obtain the feature vector corresponding to each feature value, and the feature vectors corresponding to the feature values ​​of all features can be concatenated to generate the work order feature vector of the target work order.

[0022] As an example, the target work order can be pre-sales work order 10 for ICT technology solutions. This work order can have multiple characteristics, namely: unit level, unit size, product, product type, and industry. We can obtain the feature values ​​of the unit level, unit size, product, product type, and industry characteristics of pre-sales work order 10, vectorize these feature values ​​to obtain the corresponding feature vectors, and then concatenate these feature vectors to generate the work order feature vector for pre-sales work order 10.

[0023] Among these, the characteristic values ​​for unit-level features are, for example, Grade A, Grade B, Grade C, and Grade D. The characteristic values ​​for unit size features are, for example, 100-500 people, 501-1000 people, and over 1000 people. Product characteristics include, for example, IoT security solutions and emergency communication unlocking solutions. Product type characteristics include, for example, hardware, software, and networks. Industry characteristics include, for example, mining, manufacturing, and construction.

[0024] In some embodiments, to improve the speed of obtaining work order feature vectors, a work order feature vector library can be pre-established. This library can pre-store each feature of a pre-sales work order for a technical service solution, the feature value corresponding to each feature, and the feature vector corresponding to each feature value. When obtaining the work order feature vector of a target work order, the feature values ​​of multiple features of the target work order can be obtained. Based on the feature value of each feature, the corresponding work order feature vector library for the target work order is queried. Based on the feature vector corresponding to the feature value of each queried feature, the work order feature vector is generated.

[0025] The work order feature vector library corresponding to the target work order can be the work order feature vector library of the work order type to which the target work order belongs. It can store each feature corresponding to the work order type, each feature value corresponding to each feature, and the feature vector corresponding to each feature value.

[0026] Continuing with the above example, if the target work order is a pre-sales work order 10 for ICT technology solutions, the corresponding work order feature vector library can be a work order feature vector library for pre-sales work orders in the ICT technology service solution field. This library can store the feature names, feature values, and feature value vectors of various features for different pre-sales work orders in the ICT technology service solution field. If pre-sales work order 10 includes feature 1, feature 2, and feature 3, feature value vector 1 corresponding to the feature value of feature 1, feature value vector 2 corresponding to the feature value of feature 2, and feature value vector 3 corresponding to the feature value of feature 3 can be obtained from the aforementioned work order feature vector library. Feature value vector 1, feature value vector 2, and feature value vector 3 are then concatenated to obtain the work order feature vector of pre-sales work order 10.

[0027] The candidate handlers corresponding to the target work order are those who are capable of processing the target work order. The personnel feature vector of the candidate handler is used to represent the candidate handler. It is a multi-dimensional vector, and each dimension of the vector represents a feature of the candidate handler.

[0028] In practice, for each candidate personnel, feature values ​​of multiple features of that candidate personnel can be obtained. The feature values ​​of each feature are vectorized to obtain the feature vector corresponding to the feature value of each feature. The feature vectors corresponding to the feature values ​​of each feature are concatenated together to obtain the personnel feature vector of that candidate personnel.

[0029] Among them, the candidate handlers corresponding to the target work order can be the candidate handlers corresponding to the target work order type, and the target work order type is the work order type to which the target work order belongs.

[0030] Continuing with the example above, if the target work order is pre-sales work order 10 for ICT technology solutions, the candidate handlers for the target work order can be all work order experts for pre-sales work orders of ICT technology solutions. Multiple characteristics of the candidate handlers can include: expert level, professional field, language proficiency, skills, work experience, historical work order volume, and satisfaction rating. For each candidate handler, feature values ​​of the candidate handler's expert level, professional field, language proficiency, skills, work experience, historical work order volume, and satisfaction rating can be obtained. These feature values ​​are vectorized to obtain the corresponding feature vectors. These feature vectors are then concatenated to generate the candidate handler's personnel feature vector.

[0031] Among these, the characteristic values ​​for expert level are, for example, junior, intermediate, and senior. The characteristic values ​​for professional field are, for example, hardware, software, centralized expert, and networking. The characteristic values ​​for language proficiency are, for example, Chinese, English, Japanese, and bilingual (Chinese and English). The characteristic values ​​for skill are, for example, solution development, performance optimization, penetration testing, core network, transmission network, wireless network, network security, and cloud computing. The characteristic values ​​for work experience are, for example, 1 year, 3 years, and 5 years. The characteristic values ​​for historical work orders processed are, for example, 100, 200, and 300. The characteristic values ​​for satisfaction rating are, for example, 7, 8, and 9 points.

[0032] In some embodiments, to improve the speed of acquiring personnel feature vectors, a personnel feature vector library can be pre-established. This library can store the personnel identifiers and corresponding personnel feature vectors of each work order expert in pre-sales technical service solutions within the ICT field. When acquiring the work order feature vector of a candidate processing personnel corresponding to a target work order, the library can be queried based on the candidate processing personnel's personnel identifier to obtain the corresponding personnel feature vector. The personnel feature vectors in the library can be updated as personnel characteristics change to avoid retrieving outdated personnel feature vectors.

[0033] Continuing with the above example, if the target work order is pre-sales work order 10 in the ICT technical service solution field, the personnel feature vector library corresponding to the target work order can be an expert feature vector library for work order experts in the ICT technical service solution field. This library can store the IDs of all work order experts in the ICT technical service solution field and their corresponding personnel feature vectors. If the candidate processing personnel corresponding to the target work order are work order expert 1, work order expert 2, and work order expert 3, the personnel feature vector corresponding to work order expert 1 can be obtained from the aforementioned personnel feature vector library based on work order expert 1's ID. The personnel feature vector corresponding to work order expert 2 can be obtained from the aforementioned personnel feature vector library based on work order expert 2's ID. The personnel feature vector corresponding to work order expert 3 can be obtained from the aforementioned personnel feature vector library based on work order expert 3's ID.

[0034] S104, input the work order feature vector of the target work order and the personnel feature vectors of multiple candidate processing personnel into the matching model, match the target work order with multiple candidate processing personnel, and obtain the matching degree between the target work order and each candidate processing personnel.

[0035] The matching model is a pre-trained model that takes the work order feature vector of the sample work order corresponding to the target work order and the personnel feature vector of the candidate processing personnel corresponding to the sample work order as input, and the matching degree between the sample work order and the candidate processing personnel corresponding to the sample work order as the output label, which is obtained in pre-training.

[0036] The sample work orders corresponding to the target work order can be historical work orders. The method for obtaining the work order feature vector of the sample work order is similar to that of the target work order. Similarly, the method for obtaining the personnel feature vector of the candidate processing personnel corresponding to the sample work order is similar to that of the candidate processing personnel corresponding to the target work order. In implementation, the above matching model can be trained using supervised training methods, and the weighted cross-entropy function can be used as the loss function during training.

[0037] Continuing with the example above, if the target work order is pre-sales work order 10 in the ICT technical service solution field, the corresponding sample work orders can be historical pre-sales work orders in the ICT technical service solution field, such as pre-sales work order 1, pre-sales work order 2, pre-sales work order 3, etc. The candidate handlers for these historical pre-sales work orders can be work order experts in the ICT technical service solution field. There can be multiple sample work orders, and the matching model can be iteratively trained based on the work order feature vectors of multiple sample work orders and the personnel feature vectors of the corresponding candidate handlers. During each training iteration, a historical pre-sales work order's feature vector, the corresponding work order expert's personnel feature vector, and the matching degree labels between the historical pre-sales work order and each work order expert can be obtained. The historical pre-sales work order's feature vector and the personnel feature vectors of each work order expert are input into the matching model to be trained, and the matching degree between the historical pre-sales work order and each work order expert is calculated. Then, based on the calculated matching degree between the historical pre-sales work order and each work order expert and the corresponding matching degree labels, the loss value can be calculated using the weighted cross-entropy loss function, and the model parameters of the matching model can be updated based on the loss value.

[0038] In some embodiments, to improve the model's ability to capture different features and improve matching accuracy, the matching model includes a multi-head attention model and a multi-layer perceptron model, which can be a cascaded model of the multi-head attention model and the multi-layer perceptron model. Step S104 above matches the work order feature vector of the target work order with the personnel input matching model of multiple candidate processing personnel to obtain the matching degree between the target work order and each of the candidate processing personnel. This can specifically include the following steps: The work order feature vector and the personnel feature vectors of multiple candidate handlers are input into a multi-head attention model for multi-head attention calculation to obtain a fusion vector corresponding to each candidate handler. The fusion vector corresponding to each candidate handler is then input into a multilayer perceptron model for matching degree calculation to generate a matching degree corresponding to each candidate handler.

[0039] Specifically, after receiving the work order feature vector of the target work order and the personnel feature vectors of the corresponding multiple candidate processing personnel, the multi-head attention model can perform the following processing to obtain the fusion vector corresponding to each candidate processing personnel.

[0040] Step A1: Generate a query vector based on the work order feature vector.

[0041] Specifically, this can be achieved through a linear transformation layer. This transforms the work order feature vector into a query vector. The shape of the query vector is... It can characterize the core features of the target work order.

[0042] Step A2: Generate a key vector matrix based on the skill feature vector in the personnel feature vector.

[0043] Specifically, a skill embedding layer can be used to convert the skill feature vectors in the personnel feature vectors of candidate processors into a key vector matrix Q. The shape of the key vector matrix can be... This vector can be used to capture the semantic representation of a person's skills.

[0044] Step A3: Generate a value vector matrix based on the attribute feature vectors in the personnel feature vector.

[0045] Specifically, this can be achieved through a linear transformation layer. This process converts the attribute feature vector values ​​in the personnel feature vector into a value vector matrix. The shape of the value vector matrix is... It can include attribute information such as the experience of candidate processors and user satisfaction.

[0046] Step A4: Perform multi-head attention calculation.

[0047] Specifically, a multi-head attention model can include four attention heads, each head pairing with... , , Independent weight matrices are applied to generate query, key, and value matrices for each head. A dot product is then performed on the query and key matrices for each head, followed by scaling and softmax normalization to obtain the attention weights for each head. Finally, the value matrices are weighted and summed using these attention weights to obtain the output vector for each head.

[0048] The attention weights for each head can be:

[0049] Step A5: Concatenate the outputs of each attention head to generate a fusion vector corresponding to each candidate processor.

[0050] Specifically, the output vectors of the four heads can be concatenated and then processed through a linear transformation layer. Generate final attention output The shape of the attention output can be The output integrates the correlation information of the target work order and the skills and attributes of the candidate personnel corresponding to the target work order.

[0051] The above attention output Including the fusion vectors corresponding to each processor, the attention output can be... The matching degree is calculated by inputting the multilayer perceptron model to obtain the matching degree corresponding to each processor.

[0052] A multilayer perceptron model can include an input layer, a hidden layer, and an output layer.

[0053] The input layer is used to process the work order feature vector. With attention output splicing together to form The splicing vector is the cross information of the work order requirements and the characteristics of the candidate processing personnel.

[0054] The hidden layer can contain 256 neurons. The hidden layer is used for both linear and non-linear activation based on the input data, where the activation function can be the ReLU function. Output hidden layer vector .

[0055] The output layer uses the Sigmoid function to obtain the recommendation probability for each expert, where: , , This is the weight matrix. This is a bias term.

[0056] S106. Based on the matching degree between the target work order and each candidate handler, assign handlers to the target work order.

[0057] Among these methods, the candidate handler with the highest matching degree can be identified as the handler for the target work order and assigned to the target work order.

[0058] In the above embodiments, by obtaining the work order feature vector of a target work order, such as a pre-sales work order for a technical solution, and the personnel feature vectors of multiple candidate processing personnel, the work order feature vector of the target work order is matched with the personnel input of the multiple candidate processing personnel to train a matching model, thereby obtaining the matching degree between the target work order and each of the candidate processing personnel; based on the matching degree between the target work order and each of the candidate processing personnel, processing personnel are assigned to the target work order. This enables automatic allocation of processing personnel for pre-sales work orders of technical solutions, improving the efficiency and accuracy of personnel allocation.

[0059] Furthermore, to improve the matching accuracy of the model, the matching model can be continuously optimized and updated. In some embodiments, the work order allocation method further includes the following steps: The accuracy rate of personnel allocation in the previous model cycle is calculated. If the accuracy rate is less than the accuracy threshold, the service quality information of historical work orders in the previous model cycle is obtained. The service quality information includes the work order problem resolution status and user satisfaction. Based on the service quality information of the historical work orders, the matching degree between the personnel handling the historical work orders and the historical work orders is determined. Using the work order feature vector and the personnel feature vector of the personnel handling the historical work orders as inputs, and the matching degree between the personnel handling the historical work orders and the historical work orders as the output label, the matching model of the previous model cycle is updated to obtain the matching model of the current model cycle.

[0060] Among them, accuracy can be the proportion of assigned personnel who successfully solve problems, and accuracy threshold can be the corresponding proportion threshold.

[0061] For example, the matching model can be updated weekly. If a total of [number] recommendations were made in the previous week... Several experts will handle the work orders, including If an expert successfully solved the problem, then the accuracy rate for last week can be calculated using the following formula. : .

[0062] If the accuracy threshold is 0.85, when This can trigger a model update.

[0063] When updating the model, the matching degree between the work order and the handler can be determined based on the service quality information of the work orders in the previous week. The work order feature vector and the handler feature vector of the work order in the previous week are used as inputs, and the matching degree between the handler and the work order is used as the output label to update the matching model of the previous week to obtain the matching model of the current week.

[0064] Service quality information includes: the issue resolution status of the work order and user satisfaction. The issue resolution status is, for example, resolved or unresolved, and user satisfaction is, for example, a user satisfaction rating. The matching degree between a work order and the personnel handling it can be calculated based on the following formula:

[0065] in, Indicates the degree of matching. Indicates the resolved status. This indicates the conversion of the satisfaction survey into a rating. This is a hyperparameter used to balance the importance of both.

[0066] In practice, the matching model can be updated and trained based on the following weighted cross-entropy loss function:

[0067] in, This represents the output label, i.e., the actual match degree. This indicates the degree of matching in the model's predictions.

[0068] ,

[0069] , Represents the balance coefficient. This represents a load balancing regular expression. This represents a global equilibrium constraint based on load variance. This represents a local constraint based on an overload threshold, which involves calculating the percentage of experts whose load exceeds a safe threshold and then penalizing overload behavior accordingly. Experts The current number of work orders assigned. This represents the average number of work orders for all experts. , This represents the total number of experts. This represents the threshold number of safety work orders that an expert can handle within a single cycle. This data is calculated based on historical data. , This indicates the total number of skills currently possessed by the experts. Indicating expert skills The most recent usage interval, This represents the skill freshness threshold. This penalty item sets a skill freshness threshold and counts the interval between the expert's most recent successful use of a certain skill. The longer the interval, the lower the skill freshness, and the stronger the penalty function, thereby forcing the system to prioritize assigning work orders to this type of skill.

[0070] In the above embodiments, by continuously updating and optimizing the model of the previous model period based on the service quality data of the previous model period, the model performance can be continuously improved and the accuracy of the allocation of processing personnel can be increased.

[0071] In the above embodiments, all candidate processing personnel corresponding to the target work order type are determined as candidate processing personnel corresponding to the target work order. Taking the target work order as a pre-sales work order for an ICT technical service solution, and the candidate processing personnel corresponding to the target work order as all work order experts for the pre-sales work order of the ICT technical service solution as an example, the work order allocation method is explained. In some embodiments, in order to reduce the number of matching candidate processing personnel, the candidate processing personnel corresponding to the target work order type can be screened based on the work order processing personnel screening rules to determine multiple candidate processing personnel corresponding to the target work order. The work order processing personnel screening rules may include: domain screening rules and work order load screening rules. The domain screening rules are used to screen processing personnel whose professional domain matches the domain of the user problem in the target work order, and the work order load screening rules are used to screen processing personnel whose current work order load is less than the work order load threshold. The candidate processing personnel corresponding to the target work order meet the following rules: their professional domain matches the domain of the user problem in the target work order, and their current work order load is less than the work order load threshold.

[0072] Prior to step S102 above, the work order allocation method may also include the following steps: Based on the domain of the user problem in the target work order, multiple candidate handlers corresponding to the target work order type are screened to obtain multiple first candidate handlers; wherein, the target work order type is the work order type to which the target work order belongs; the first candidate handlers are those whose professional fields match the domain of the user problem among the multiple candidate handlers corresponding to the target work order type; based on the current work order load of the first candidate handlers and their corresponding personal work order load threshold, the multiple first candidate handlers are screened to obtain multiple second candidate handlers; the second candidate handlers are those whose current work order load is less than their corresponding personal work order load threshold among the multiple first candidate handlers; the multiple second candidate handlers are determined as the multiple candidate handlers corresponding to the target work order.

[0073] Taking pre-sales work order 10, which targets ICT technical solutions, as an example, if the user's problem in the target work order is a 5G network optimization issue, and the problem falls under the "communication field," experts with expertise in the "communication field" can be selected from the pre-sales work order experts in ICT technical solutions. Then, the expert's current work order load and personal work order load threshold can be obtained. If the expert's current work order load is less than their personal work order load threshold, that expert can be identified as a candidate handler for the target work order. Here, the current work order load is, for example, the number of work orders the expert currently has not processed, and the personal work order load threshold is, for example, the maximum number of work orders the expert can process.

[0074] In the above embodiments, by initially screening candidate personnel corresponding to the target work order based on domain filtering conditions and work order workload filtering conditions, the number of candidate personnel to be matched with the target work order can be reduced, thereby improving the efficiency of subsequent matching.

[0075] In the above embodiments, the individual work order load threshold is a personalized load threshold for the processing personnel, representing the maximum work order load that the processing personnel can handle. Different processing personnel may have different individual work order load thresholds. By selecting processing personnel based on individual work order load thresholds, it can be ensured that the work order load of the processing personnel matches the work order load they can handle, avoiding overworking of processing personnel and affecting processing quality.

[0076] In implementation, the individual work order load threshold for each processor can be updated periodically. The above work order allocation method may also include the following steps: Based on the efficiency information and service quality information of the candidate processing personnel in the previous threshold period, the efficiency information and business fluctuation information of the aforementioned candidate processing personnel in the current threshold period, and the level information of the aforementioned candidate processing personnel, the individual work order load threshold of the aforementioned candidate processing personnel in the current threshold period is determined.

[0077] The individual work order load threshold can be updated at a specified time within the current threshold period. For example, the threshold period can be one day, and the individual work order threshold can be updated every day at noon. At 12:00 noon each day, the individual work order threshold for each handler is determined and updated based on the handler's efficiency and service quality information from the previous day, their processing efficiency information for the current day, their business fluctuation information for the day, and their level information. When filtering candidate handlers based on the individual work order threshold, the latest individual work order threshold for each candidate handler at the time of filtering can be obtained and used for filtering.

[0078] The efficiency information for personnel handling tasks in the previous threshold period may include the average processing time per task for that personnel in that period. The service quality information for the previous threshold period may include the user satisfaction rating and problem resolution rate for that personnel in that period. The efficiency information for personnel handling tasks in the current threshold period may include the time spent processing tasks by that personnel in the current threshold period. The business fluctuation information for the current threshold period may include the year-on-year growth rate of task volume in the current threshold period compared to the same period of the previous threshold period, and whether the current period is a peak business period. Personnel level information may include, for example, junior, intermediate, and senior levels.

[0079] In implementation, the current work order load of a candidate handler can be the number of currently unprocessed work orders, and their individual work order load threshold can be the maximum number of work orders they can process. The individual work order load threshold for this candidate handler in the current threshold period can be calculated based on the following formula:

[0080] in, Indicates the baseline work order quantity. This indicates the individual work order load threshold. Indicates the efficiency factor weight. Represents the efficiency factor. Indicates the weight of the quality factor. Representative service quality factor, Represents the weight of the business volatility factor. Represents business volatility factor, Representative level weighting Representative level factor.

[0081] efficiency factor The average processing time per work order can be based on the work orders already processed by candidate processors in the previous threshold period. Compared with the average processing time of work orders in this threshold period It is certain that the calculation formula can be: .

[0082] As an example, if the threshold period is one day, and the individual work order threshold is updated every noon, the efficiency factor of the current candidate worker can be determined based on the average time taken to process work orders handled by the candidate worker the previous day and the average time taken to process work orders handled by the candidate worker that morning. By comparing the efficiency of the current day with historical data, the current working status of the candidate worker can be reflected in real time. If the expert's processing speed increases on the current day, the efficiency factor increases, and the threshold is raised accordingly, which can fully realize the potential of the candidate worker to work efficiently. If the efficiency decreases, the threshold is lowered, which can prevent the candidate worker from being overworked.

[0083] quality factor It can be based on the average user satisfaction score of candidate users in the previous threshold period. With problem resolution rate The weighted average calculation can be performed using the following formula: .

[0084] in, The weight is used to emphasize the importance of service quality to load allocation. Only when experts achieve high levels in both satisfaction and resolution rate (e.g., satisfaction ≥ 4.5 points and resolution rate ≥ 95%) will the quality factor be at a high level, and the threshold will increase accordingly. This forms a positive incentive mechanism that "high service quality means more work orders can be taken," ensuring service quality while improving user satisfaction.

[0085] Using the above example, if the threshold period is one day and the individual work order threshold is updated every noon, the quality factor of the candidate worker can be determined based on the average user satisfaction rating and problem resolution rate of the work orders handled by the candidate worker on the previous day.

[0086] Business volatility factor The growth rate of work order volume in the current threshold period compared to the same period of the previous threshold period can be used as a basis. and business sudden fluctuation coefficient The calculation formula can be: .

[0087] Among them, during peak business periods, the business volatility coefficient is... If it is not during peak business hours, that is, during regular business hours.

[0088] Using the above example, if the threshold period is one day and the individual work order threshold is updated every noon, the ratio of the number of work orders processed by the candidate processing personnel at noon of the day to the number of work orders processed by the candidate processing personnel at noon of the previous day can be determined as the growth rate. Based on this growth rate and whether the day is a peak business period, the business fluctuation factor for the day can be determined.

[0089] Personnel level factor The level of the candidate handler can be used as the determining factor. Different levels have different level factors. For example, if the candidate handler's level is junior, the level factor is 0.8; if the candidate handler's level is intermediate, the level factor is 1.0; and if the candidate handler's level is senior, the level factor can be 1.2. By setting different level factors for different levels of personnel, and setting a lower threshold for junior personnel, the system can ensure that work order workload allocation is tilted towards highly skilled experts, thereby optimizing work order processing efficiency and quality.

[0090] In the above embodiments, the target work order can be a work order without an scheduled processing time. Based on the above filtering method, candidate handlers can be filtered for the target work order in real time, and then handlers can be assigned to the target work order in real time. In some embodiments, the target work order can be a work order with an scheduled processing time. Before step S102, the above work order assignment method may include the following steps: Based on the domain of the user's problem in the target work order, multiple candidate handlers corresponding to the target work order type are screened to obtain multiple first candidate handlers; wherein, the target work order type refers to the work order type to which the target work order belongs; the first candidate handlers are those whose professional domain matches the domain of the user's problem among the multiple candidate handlers corresponding to the target work order type; based on the current work order load of the first candidate handlers and their corresponding personal work order load threshold, the multiple first candidate handlers are screened to obtain multiple second candidate handlers; the second candidate handlers are those whose current work order load is less than their corresponding personal work order load threshold among the multiple first candidate handlers; based on the user's scheduled time period in the target work order and the candidate handlers, the multiple second candidate handlers are screened to obtain multiple third candidate handlers, wherein the third candidate handlers are those whose idle time matches the user's scheduled time period among the multiple second candidate handlers.

[0091] For example, if the user's appointment time in the target work order is from 2:00 PM to 3:00 PM on the same day, personnel who are available from the second candidate handlers during the same time period can be selected as the third candidate handlers.

[0092] In the above embodiments, by initially screening candidate personnel corresponding to the target work order based on domain filtering conditions, work order load filtering conditions, and time filtering adjustment, the number of candidates to be matched with the target work order can be reduced, thereby improving the efficiency of subsequent matching.

[0093] Corresponding to the work order allocation method provided in the above embodiments, based on the same technical concept, this application also provides a work order allocation device. Figure 2 This is a schematic diagram of the work order allocation device according to some embodiments of this application. For example... Figure 2 As shown, in some embodiments of this application, the work order allocation device 200 includes: an acquisition module 210, a matching module 220, and an allocation module 230.

[0094] The acquisition module 210 is used to acquire the work order feature vector of the target work order and the personnel feature vectors of multiple candidate processing personnel corresponding to the target work order.

[0095] The matching module 220 is used to match the target work order's work order feature vector with the personnel input matching model of the multiple candidate processing personnel, and obtain the matching degree between the target work order and each candidate processing personnel; wherein, the matching model is pre-trained by taking the work order feature vector of the sample work order corresponding to the target work order and the personnel feature vector of the candidate processing personnel corresponding to the sample work order as input, and the matching degree label between the sample work order and the candidate processing personnel corresponding to the sample work order as output label.

[0096] The allocation module 230 is used to allocate processing personnel to the target work order based on the matching degree between the target work order and each of the candidate processing personnel.

[0097] In some embodiments of this application, the matching model includes: a multi-head attention model and a multilayer perceptron model; the matching module 220 is specifically used for: The work order feature vector of the target work order and the personnel feature vectors of the multiple candidate processing personnel are input into the multi-head attention model for multi-head attention calculation to obtain the fusion vector corresponding to each candidate processing personnel. The fusion vector corresponding to each candidate processor is input into the multilayer perceptron model to calculate the matching degree, thus obtaining the matching degree corresponding to each candidate processor.

[0098] In some embodiments of this application, the matching model is updated periodically, and the apparatus further includes: The model update module is used to calculate the allocation accuracy of the processing personnel in the previous model cycle; if the allocation accuracy is less than the accuracy threshold, it obtains the service quality information of historical work orders in the previous model cycle; the service quality information includes: work order problem resolution status and user satisfaction; based on the service quality information of the historical work orders, it determines the matching degree between the processing personnel of the historical work orders and the historical work orders; using the work order feature vector and the personnel feature vector of the processing personnel of the historical work orders as inputs, and the matching degree between the processing personnel of the historical work orders and the historical work orders as the output label, it updates the matching model of the previous model cycle to obtain the matching model of the current model cycle.

[0099] In some embodiments of this application, the acquisition module 210 is specifically used for: The feature values ​​of multiple features of the target work order are obtained. Based on the feature value of each feature, the work order feature vector library corresponding to the target work order is queried. Based on the feature vector corresponding to the feature value of each feature, the work order feature vector is generated. Based on the personnel identifier of each candidate processing personnel, the personnel feature vector library corresponding to the target work order is queried to obtain the personnel feature vector of each candidate processing personnel.

[0100] In some embodiments of this application, the apparatus further includes: The candidate handler determination module is used to filter multiple candidate handlers corresponding to the target work order type based on the domain of the user problem in the target work order, thereby obtaining multiple first candidate handlers; wherein, the target work order type is the work order type to which the target work order belongs; the first candidate handlers are those whose professional field matches the domain of the user problem among the multiple candidate handlers corresponding to the target work order type; based on the current work order load of the first candidate handlers and their corresponding personal work order load threshold, the multiple first candidate handlers are filtered to obtain multiple second candidate handlers; the first candidate handlers are those whose current work order load is less than their corresponding personal work order load threshold among the multiple second candidate handlers; and the multiple second candidate handlers are determined as the multiple candidate handlers corresponding to the target work order.

[0101] In some embodiments of this application, the individual work order load threshold is updated periodically, and the device further includes: The threshold update module is used to determine the individual work order load threshold of the candidate processing personnel in the current threshold period based on the efficiency information, service quality information, efficiency information, business fluctuation information, and level information of the candidate processing personnel in the previous threshold period.

[0102] The work order allocation device provided in this application embodiment can realize the various processes in the above work order allocation method embodiment, and will not be described again here to avoid repetition.

[0103] Corresponding to the work order allocation method provided in the above embodiments, based on the same technical concept, this application also provides a work order allocation device.

[0104] When using integrated units, electronic devices may include processing modules and storage modules.

[0105] The processing module may be a processor or a controller, which can implement or execute various exemplary logic blocks, modules, and circuits as disclosed in this application. The processor may also be a combination of computing functions, such as a combination of one or more microprocessors, a combination of digital signal processing (DSP) and a microprocessor, etc., and the storage module may be a memory.

[0106] In some embodiments, the electronic device may include a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the work order allocation method described above. To avoid repetition, these will not be elaborated further here.

[0107] Furthermore, corresponding to the work order allocation method provided in the above embodiments, based on the same technical concept, this application also provides a computer-readable storage medium storing a computer program thereon, the computer program being executable by a processor to implement the steps of the above work order allocation method. To avoid repetition, further details are omitted here.

[0108] Furthermore, corresponding to the work order allocation method provided in the above embodiments, based on the same technical concept, this application also provides a computer program product, which includes a computer program that can be executed by a processor to implement the steps of the above work order allocation method. To avoid repetition, further details are omitted here.

[0109] The above describes the work order allocation method, system, device, readable storage medium, and computer program product according to embodiments of this application.

[0110] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus, readable storage media, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0111] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0112] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0113] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0114] In a typical configuration, an electronic device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0115] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0116] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0117] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0118] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0119] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A work order allocation method, characterized in that, The method includes: Obtain the work order feature vector of the target work order and the personnel feature vectors of multiple candidate processing personnel corresponding to the target work order; The work order feature vector of the target work order is matched with the personnel input matching model of the multiple candidate processing personnel to obtain the matching degree between the target work order and each candidate processing personnel; wherein, the matching model is pre-trained by taking the work order feature vector of the sample work order corresponding to the target work order and the personnel feature vector of the candidate processing personnel corresponding to the sample work order as input, and taking the matching degree label between the sample work order and the candidate processing personnel corresponding to the sample work order as output label; Based on the matching degree between the target work order and each of the candidate processing personnel, a processing personnel is assigned to the target work order.

2. The method according to claim 1, characterized in that, The matching model includes a multi-head attention model and a multilayer perceptron model; the step of matching the work order feature vector of the target work order with the personnel input matching model of the multiple candidate processing personnel to match the target work order with the multiple candidate processing personnel and obtain the matching degree between the target work order and each candidate processing personnel includes: The work order feature vector of the target work order and the personnel feature vectors of the multiple candidate processing personnel are input into the multi-head attention model for multi-head attention calculation to obtain the fusion vector corresponding to each candidate processing personnel. The fusion vector corresponding to each candidate processor is input into the multilayer perceptron model to calculate the matching degree, thus obtaining the matching degree corresponding to each candidate processor.

3. The method according to claim 1, characterized in that, The matching model is updated periodically, and the method further includes: The accuracy rate of personnel allocation in the previous model cycle was statistically analyzed. If the allocation accuracy is less than the accuracy threshold, obtain the service quality information of historical work orders from the previous model cycle; the service quality information includes: work order problem resolution status and user satisfaction. The matching degree between the personnel handling the historical work orders and the historical work orders is determined based on the service quality information of the historical work orders. Using the work order feature vector of the historical work order and the personnel feature vector of the personnel who processed the historical work order as input, and the matching degree between the personnel who processed the historical work order and the historical work order as the output label, the matching model of the previous model cycle is updated to obtain the matching model of the current model cycle.

4. The method according to claim 1, characterized in that, The process of obtaining the work order feature vector of the target work order and the personnel corresponding to the target work order and multiple candidate processing personnel includes: Obtain feature values ​​of multiple features of the target work order, query the work order feature vector library corresponding to the target work order based on the feature value of each feature, and generate the work order feature vector based on the feature vector corresponding to the feature value of each feature found in the query. Based on the personnel identifier of each candidate processing personnel, the personnel feature vector library corresponding to the target work order is queried to obtain the personnel feature vector of each candidate processing personnel.

5. The method according to claim 1, characterized in that, Before obtaining the personnel feature vectors of multiple candidate processing personnel corresponding to the target work order, the method further includes: Based on the domain of the user's problem in the target work order, multiple candidate handlers corresponding to the target work order type are screened to obtain multiple first candidate handlers; wherein, the target work order type is the work order type to which the target work order belongs; the first candidate handlers are the personnel whose professional field matches the domain of the user's problem among the multiple candidate handlers corresponding to the target work order type. Based on the current work order load of the first candidate handler and its corresponding personal work order load threshold, the plurality of first candidate handlers are screened to obtain a plurality of second candidate handlers; the second candidate handlers are those among the plurality of first candidate handlers whose current work order load is less than their corresponding personal work order load threshold. Based on the multiple second candidate handlers, multiple candidate handlers corresponding to the target work order are determined.

6. The method according to claim 5, characterized in that, The individual work order load threshold is updated periodically, and the method further includes: Based on the efficiency information and service quality information of the candidate processing personnel in the previous threshold period, the efficiency information and business fluctuation information of the candidate processing personnel in this threshold period, and the level information of the candidate processing personnel, the individual work order load threshold of the candidate processing personnel in this threshold period is determined.

7. A work order distribution device, characterized in that, The device includes: The acquisition module is used to acquire the work order feature vector of the target work order and the personnel feature vectors of multiple candidate processing personnel corresponding to the target work order. The matching module is used to match the work order feature vector of the target work order with the personnel input matching model of the multiple candidate processing personnel, and obtain the matching degree between the target work order and each candidate processing personnel; wherein, the matching model is pre-trained by taking the work order feature vector of the sample work order corresponding to the target work order and the personnel feature vector of the candidate processing personnel corresponding to the sample work order as input, and the matching degree label between the sample work order and the candidate processing personnel corresponding to the sample work order as output label; The allocation module is used to allocate processing personnel to the target work order based on the matching degree between the target work order and each of the candidate processing personnel.

8. A work order distribution device, characterized in that, The device includes: A memory on which computer programs are stored; A processor for executing the computer program to implement the steps of the work order allocation method according to any one of claims 1 to 6.

9. A readable storage medium, characterized in that, The storage medium stores a computer program that can be executed by a processor to implement the steps of the work order allocation method according to any one of claims 1 to 6.

10. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the steps of the work order allocation method according to any one of claims 1 to 6.

Images