Work order allocation method and device based on deep collaborative filtering and related equipment

Abstract

The application discloses a work order distribution method and device based on deep collaborative filtering and related equipment. The method comprises the following steps: obtaining historical data of time required by workers to complete work orders, and preprocessing the historical data to obtain sample data; inputting the sample data into a preset deep collaborative filtering model for time prediction training to obtain a target collaborative filtering model; inputting a to-be-distributed work order into the target collaborative filtering model to predict the completion time of the to-be-distributed work order completed by each worker, and obtaining a prediction time matrix; batch dividing the to-be-distributed work order according to regions to obtain a plurality of small batches of work orders; performing minimum cost calculation on the prediction time matrix corresponding to each small batch of work orders by using a Hungarian algorithm to obtain an optimal distribution matrix of each small batch of work orders, and distributing the work order according to the optimal distribution matrix. The method realizes optimal distribution of work orders, improves work order delivery efficiency, and achieves the purpose of cost reduction and benefit increase of enterprises.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of work order distribution, and in particular to a work order distribution method and device based on deep collaborative filtering and related equipment. BACKGROUND

[0002] The work order dispatch system of the current online car-hailing is crucial to improve the production efficiency and save the cost of the enterprise. The work order dispatch system dispatches the most suitable work order for the worker according to the working ability of the worker, the adaptability to different work and the historical work order status, thereby reducing the total work order time or total cost. The work order dispatch system with high efficiency can greatly save the time and cost of the worker to complete the work order, thereby improving the profit of the enterprise operation.

[0003] At present, most of the work order dispatch systems of online car-hailing separately use operational optimization algorithm or collaborative filtering algorithm for work order dispatch optimization. The purpose of the operational optimization algorithm is to use mathematical methods to study the optimization approach and scheme for the studied system, to obtain a reasonable best scheme of using manpower, material resources and financial resources, and to improve the efficiency and benefit of the system, thereby providing a scientific basis for decision makers. The collaborative filtering algorithm finds a certain similarity matrix (similarity between users or similarity between objects) through the behavior of the group, and makes decisions and recommendations for users through the similarity matrix. The deep collaborative filtering algorithm is a technology that uses a neural network to replace the similarity matrix decomposition part to learn the arbitrary function relationship between user and object hidden features. However, the optimization effect is poor and the user experience is poor when the operational optimization algorithm or the collaborative filtering algorithm is separately used for work order dispatch optimization. SUMMARY

[0004] Embodiments of the present application provide a work order distribution method and device based on deep collaborative filtering and related equipment, aiming at solving the problem of poor work order dispatch optimization effect in the prior art.

[0005] In a first aspect, embodiments of the present application provide a work order distribution method based on deep collaborative filtering, which comprises:

[0006] obtaining historical data of the time required for a worker to complete a work order, and preprocessing the historical data to obtain sample data, wherein the historical data comprises worker information, historical work order and work order completion time;

[0007] inputting the sample data into a preset deep collaborative filtering model for time prediction training to obtain a target collaborative filtering model;

[0008] inputting a to-be-assigned work order into the target collaborative filtering model to predict the completion time of the to-be-assigned work order for each worker, and obtaining a prediction time matrix;

[0009] The to-be-assigned work orders are batched according to regions to obtain a plurality of small-batch work orders;

[0010] The optimal assignment matrix of each small-batch work order is obtained by performing minimum cost calculation on the prediction time matrix corresponding to each small-batch work order through the Hungarian algorithm, and work order assignment is performed according to the optimal assignment matrix.

[0011] In a second aspect, an embodiment of the present application provides a work order assignment device based on deep collaborative filtering, comprising:

[0012] The acquisition module is configured to acquire historical data of time required by workers to complete work orders, and pre-process the historical data to obtain sample data, wherein the historical data comprises worker information, historical work orders and work order completion time.

[0013] The training module is configured to input the sample data into a preset deep collaborative filtering model to perform time prediction training, and obtain a target collaborative filtering model.

[0014] The prediction module is configured to input to-be-assigned work orders into the target collaborative filtering model to predict completion time of the to-be-assigned work orders by each worker, and obtain a prediction time matrix.

[0015] The batch module is configured to batch the to-be-assigned work orders according to regions to obtain a plurality of small-batch work orders.

[0016] The optimization module is configured to perform minimum cost calculation on the prediction time matrix corresponding to each small-batch work order through the Hungarian algorithm, to obtain an optimal assignment matrix of each small-batch work order, and perform work order assignment according to the optimal assignment matrix.

[0017] In a third aspect, an embodiment of the present application further provides a computer device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the work order assignment method based on deep collaborative filtering of the first aspect when executing the computer program.

[0018] In a fourth aspect, an embodiment of the present application further provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program causes a processor to execute the work order assignment method based on deep collaborative filtering of the first aspect when the computer program is executed by the processor.

[0019] The embodiment of the present application provides a work order distribution method and device based on deep collaborative filtering and related equipment. The method comprises the following steps: obtaining historical data of time required by workers to complete work orders, and preprocessing the historical data to obtain sample data; inputting the sample data into a preset deep collaborative filtering model for time prediction training to obtain a target collaborative filtering model; inputting a to-be-distributed work order into the target collaborative filtering model to predict the completion time of the to-be-distributed work order by each worker, and obtaining a prediction time matrix; batch dividing the to-be-distributed work order according to regions to obtain a plurality of small batches of work orders; performing minimum cost calculation on the prediction time matrix corresponding to each small batch of work orders by using a Hungarian algorithm to obtain an optimal distribution matrix of each small batch of work orders, and distributing the work order according to the optimal distribution matrix. The method predicts the completion time of the to-be-distributed work order based on the deep collaborative filtering model, optimizes the prediction result by combining batch processing and the Hungarian algorithm, obtains the best distribution matrix of the to-be-distributed work order, optimally distributes the work order, improves the work order delivery efficiency, and achieves the purpose of cost reduction and benefit increase of an enterprise. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0021] Figure 1 The flowchart of an embodiment of the work order distribution method based on deep collaborative filtering provided by the present application;

[0022] Figure 2 The structural diagram of the deep collaborative filtering model in the work order distribution method based on deep collaborative filtering provided by the present application;

[0023] Figure 3 The schematic block diagram of the work order distribution device based on deep collaborative filtering provided by the present application;

[0024] Figure 4 The schematic diagram of batch processing of the prediction time matrix in the work order based on deep collaborative filtering provided by the present application. DETAILED DESCRIPTION

[0025] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application.

[0026] It should be understood that the terms "comprises" and "comprising," when used in this specification and the following claims, indicate the presence of the described features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0027] It should also be understood that the terms used in the specification of the application herein are used for the purpose of describing particular embodiments only and are not intended to limit the application. As used in the specification and the appended claims of the application, the singular forms "a," "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

[0028] It should be further understood that the term "and / or" used in the specification of the application and the appended claims means one or more of the associated listed items as well as all possible combinations of the items and includes these combinations.

[0029] Please refer to Figure 1 shown, Figure 1 A flowchart of a deep collaborative filtering-based work order allocation method provided by an embodiment of the application is shown in the figure. The method comprises steps S110-S150.

[0030] In step S110, historical data of time required by workers to complete work orders is acquired, and the historical data is preprocessed to obtain sample data. The historical data comprises worker information, historical work orders and work order completion time.

[0031] In this embodiment, the historical data of time required by workers to complete work orders is acquired from a work order dispatching system of a network car-hailing service. The historical data comprises worker information, historical work orders and work order completion time. The historical data can be acquired from a database or downloaded from an online source. In addition, the worker information comprises worker ID, worker mobile phone number, worker age, worker evaluation, worker post and worker location latitude and longitude. The information of the historical work orders in the historical data further comprises order ID, order location latitude and longitude, order priority and order type. Since the time required to complete different work orders can be quite different, which can affect model training, the historical data is preprocessed to obtain sample data.

[0032] In an embodiment, the preprocessing of the historical data comprises:

[0033] In step one, the historical work order matrix corresponding to the historical data is standardized to obtain standard data.

[0034] In step two, the standard data is subjected to anomaly detection to obtain anomaly data.

[0035] In step three, the anomaly data is removed to obtain normal data.

[0036] Step four, missing value filling and data conversion are performed on the normal data to obtain the sample data.

[0037] In this embodiment, in order to improve the quality of historical data, the historical data is preprocessed, specifically including: first, because the time required to complete different work orders may be greatly different, which will affect the convergence speed of the model, in order to eliminate this influence, 0-1 normalization technology is used to standardize the historical work order matrix, and the work order completion time is scaled to the interval [0, 1], the historical work order matrix corresponding to the historical data is standardized according to the following formula, and the standard data is obtained:

[0038]

[0039] wherein, represents the standard value, and θ ij represents the real completion time of worker j to complete work order i, min represents the minimum time required for all workers to complete work order i, and max represents the maximum time required for all workers to complete work order i;

[0040] Then, Z-score method is used for outlier detection, assuming that the sample number is x, the average value of all samples is x m , and s is the standard deviation of all samples, then

[0041] Z = (x-x m ) / s,

[0042] If the absolute value of Z of a sample is greater than 3, the sample is abnormal data; the abnormal data is removed to obtain normal data. Because the worker age and worker evaluation have missing values, the mean filling method is used to fill the missing values. Finally, the order priority, order type, worker age, worker evaluation and worker post are converted into category features to obtain the sample data.

[0043] Step S120, inputting the sample data into the preset deep collaborative filtering model for time prediction training to obtain a target collaborative filtering model;

[0044] In this embodiment, the sample data is input into the preset deep collaborative filtering model for training, for example, Figure 2As shown, specifically includes: respectively on the sample data of worker information and historical work order one-hot encoding, get worker information and historical work order corresponding binary sparse features (namely user features and item features); then through the embedding layer (embedding layer is full connection layer) of deep collaborative filtering model to binary sparse feature mapping, get worker information and historical work order dense vector, the dense vector obtained can be regarded as the hidden vector of the implicit factor model (user hidden vector and item hidden vector); then the dense vector is input into the deep collaborative filtering layer of the deep collaborative filtering model for vector mapping, to obtain the initial prediction time (i.e. predicted value); based on the initial prediction time and the actual time of the work order completion (target value), according to the preset loss function, the time prediction loss is calculated, and the time prediction loss is obtained; based on the time prediction loss, the model parameters of the deep collaborative filtering model are optimized, and the target collaborative filtering model is obtained. Wherein, the work order completion time is predicted as follows:

[0045]

[0046] Wherein, θ ij represents the initial prediction time of worker j completing work order i, represents the vector composed of all workers, represents the vector composed of all work orders, P and Q represent the hidden feature matrix of workers and work orders respectively, P T and Q T represent the rotation matrix of the hidden feature matrix of workers and work orders respectively, Θ f represents the model parameters of the multi-layer neural network f;

[0047] The formula can be further expressed as:

[0048]

[0049] Wherein, φ out represents the mapping function of the output layer, φ X represents the Xth network layer.

[0050] Further, after the prediction of the deep collaborative filtering model is completed, since the work order completion time is scaled to the [0, 1] interval in the aforementioned preprocessing of historical data, in order to calculate the time prediction loss, the initial prediction time needs to be converted back to the time value, so the initial prediction time is converted back to the real value according to the following formula, which is the time value:

[0051] θ true = (max-min) * θ pred +min;

[0052] Wherein, θ predθi represents the initial predicted time for a worker to complete a work order i after standardization true θi represents the actual time for a worker to complete a work order i after mapping, min represents the minimum time required for a worker to complete a work order i, and max represents the maximum time required for all workers to complete a work order i.

[0053] Step S130, input the to-be-assigned work order into the target collaborative filtering model to predict the completion time of each worker completing the to-be-assigned work order, and obtain a predicted time matrix;

[0054] Step S140, batch divide the to-be-assigned work order according to regions to obtain a plurality of small batch work orders;

[0055] Step S150, perform minimum cost calculation on the predicted time matrix corresponding to each small batch work order by using the Hungarian algorithm to obtain an optimal assignment matrix of each small batch work order, and perform work order assignment according to the optimal assignment matrix.

[0056] In this embodiment, the to-be-assigned work order is input into the target collaborative filtering model to predict the completion time of each worker completing the to-be-assigned work order, and a predicted time matrix of the to-be-assigned work order is obtained. Then, the Hungarian algorithm is used to optimize the predicted time matrix, but the model running time increases exponentially with the increase of the number of samples due to the poor scalability of the Hungarian algorithm. In order to overcome this shortcoming, the to-be-assigned work order is processed in small batches, specifically: Figure 4 As shown in FIG. 1, the to-be-assigned work order is batch divided according to regions, and the to-be-assigned work order is split into a plurality of small batch work orders (Batch1,..., BatchN), and then the Hungarian algorithm is used to perform minimum cost calculation on the predicted time matrix corresponding to each small batch work order to obtain an optimal assignment matrix of each small batch work order. Wherein, Job1, Job m , Job n and Job k represent work orders, worker1, worker p , worker q and worker w represent workers, score11, scorew1, score1k, scorewk, scorep1, score1m, scorepm, scoreqn, scorewn, scoreqk and scorewk represent standard values. After obtaining the optimal assignment matrix of each small batch work order, the standard values in the optimal assignment matrix are converted back to real values, and then it is determined whether to assign the work order to the worker according to the real values. Wherein, the work order is assigned according to the real values of the standard values in the optimal assignment matrix to achieve optimal assignment of the to-be-assigned work order. The optimal assignment matrix is calculated according to the following formula:

[0057]

[0058] Wherein, W, K respectively represent workers and work order set, φ ij Indicates the predicted time of worker j completing work order i, z ij Indicates whether work order i is assigned to worker j, if yes, the value is 1, otherwise the value is 0.

[0059] Further, considering the continuity of work order before and after, that is, the case that workers complete one work order and then another work order, the application proposes to bundle orders with similar distances, set a distance threshold S, if the distance of two work orders is less than S, they are divided into the same work order for unified distribution, thereby the number of workers can be reduced to reduce the cost of order distribution.

[0060] The method predicts the completion time of the to-be-assigned work order based on a deep collaborative filtering model, optimizes the prediction result by combining batch processing and the Hungarian algorithm, obtains the optimal assignment matrix of the to-be-assigned work order, realizes optimal assignment of the work order, improves the work order delivery efficiency, and achieves the purpose of reducing cost and increasing efficiency of the enterprise.

[0061] The embodiment of the application also provides a work order assignment device based on deep collaborative filtering, which is used to execute any one of the foregoing embodiments of the work order assignment method based on deep collaborative filtering. Figure 3 , Figure 3 is a schematic block diagram of the work order assignment device based on deep collaborative filtering provided by the embodiment of the application. The work order assignment device based on deep collaborative filtering 100 can be configured in a service end node.

[0062] As shown in Figure 3 , the work order assignment device based on deep collaborative filtering 100 includes an acquisition module 110, a training module 120, a prediction module 130, a batch processing module 140, and an optimization module 150.

[0063] The acquisition module 110 is used to acquire historical data of time required by workers to complete work orders, and pre-process the historical data to obtain sample data, wherein the historical data includes worker information, historical work orders, and work order completion time.

[0064] The training module 120 is used to input the sample data into a preset deep collaborative filtering model to perform time prediction training, and obtain a target collaborative filtering model.

[0065] The prediction module 130 is used to input a to-be-assigned work order into the target collaborative filtering model to predict the completion time of the to-be-assigned work order completed by each worker, and obtain a predicted time matrix.

[0066] The batch module 140 is configured to divide the to-be-allocated work orders into small batches according to regions, to obtain a plurality of small-batch work orders.

[0067] The optimization module 150 is configured to perform minimum cost calculation on the predicted time matrix corresponding to each small-batch work order by using a Hungarian algorithm, to obtain an optimal allocation matrix of each small-batch work order, and to perform work order allocation according to the optimal allocation matrix.

[0068] In an embodiment, the acquisition module 110 includes:

[0069] The standardization unit is configured to perform standardization processing on the historical work order matrix corresponding to the historical data, to obtain standard data.

[0070] The anomaly detection unit is configured to perform anomaly detection on the standard data, to obtain anomaly data.

[0071] The elimination unit is configured to eliminate the anomaly data, to obtain normal data.

[0072] The conversion unit is configured to perform missing value filling and data conversion on the normal data, to obtain the sample data.

[0073] In an embodiment, the training module 120 includes:

[0074] The encoding unit is configured to perform one-hot encoding on the worker information and the historical work order of the sample data respectively, to obtain binary sparse features corresponding to the worker information and the historical work order.

[0075] The first mapping unit is configured to map the binary sparse features through a fully connected layer of the deep collaborative filtering model, to obtain a dense vector.

[0076] The second mapping unit is configured to input the dense vector into a deep collaborative filtering layer of the deep collaborative filtering model for vector mapping, to obtain an initial predicted time.

[0077] The loss calculation unit is configured to calculate a time prediction loss according to a preset loss function based on the initial predicted time and an actual time of work order completion, to obtain the time prediction loss.

[0078] The parameter optimization unit is configured to optimize model parameters of the deep collaborative filtering model based on the time prediction loss, to obtain a target collaborative filtering model.

[0079] The embodiment of the present application also provides a computer device, including a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the deep collaborative filtering-based work order allocation method as described above when executing the computer program.

[0080] In another embodiment of the present application, a computer readable storage medium is provided. The computer readable storage medium can be a non-transitory computer readable storage medium. The computer readable storage medium stores a computer program which, when executed by a processor, causes the processor to perform the method for ticket assignment based on deep collaborative filtering as described above.

[0081] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices, apparatuses and units can refer to the corresponding processes in the foregoing method embodiments, which will not be described here. Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, each example has been described in the above description in a general manner in terms of its functionality. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0082] In several embodiments provided by the present application, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For example, the division of the units is merely a logical function division, and actual implementation can have another division manner, or units with the same function can be combined into one unit, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interfaces, devices or units, and can also be electrical, mechanical or other forms of connection.

[0083] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, i.e. they can be located in one place or distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment of the present application.

[0084] In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically, or two or more units can be integrated into one unit. The integrated unit can be realized in the form of hardware or software functional unit.

[0085] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a storage medium. Based on such understanding, the technical solutions of the present application essentially or say the part that contributes to the prior art, or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server node, or a network device, etc.) to execute all or part of the steps of the method described in various embodiments of the present application. The aforementioned storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0086] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present application, and these modifications or replacements should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for ticket assignment based on deep collaborative filtering, characterized in that, The method comprises the following steps: obtaining historical data of time required by workers to complete work orders, and preprocessing the historical data to obtain sample data, wherein the historical data comprises worker information, historical work orders and work order completion time; inputting the sample data into a preset deep collaborative filtering model for time prediction training to obtain a target collaborative filtering model; inputting a to-be-assigned work order into the target collaborative filtering model to predict the completion time of the to-be-assigned work order by each worker, and obtaining a predicted time matrix; dividing the to-be-assigned work order into multiple small batches according to regions to obtain multiple small batches of work orders; performing minimum cost calculation on the predicted time matrix corresponding to each small batch of work orders by using the Hungarian algorithm to obtain an optimal assignment matrix of each small batch of work orders, and performing work order assignment according to the optimal assignment matrix; the preprocessing of the historical data to obtain sample data comprises: performing standardization processing on a historical work order matrix corresponding to the historical data to obtain standard data; performing anomaly detection on the standard data to obtain abnormal data; removing the abnormal data to obtain normal data; and performing missing value filling and data conversion on the normal data to obtain the sample data; wherein the anomaly detection on the standard data is performed according to the following formula to obtain abnormal values: Z = (x - x m ) / s Wherein, Z represents the abnormal value, x represents the sample number, x m represents the average value of all samples, s represents the standard deviation of all samples; if the absolute value of Z of a sample is greater than 3, the sample is the abnormal data; the inputting of the sample data into a preset deep collaborative filtering model for time prediction training to obtain a target collaborative filtering model comprises: one-hot encoding the worker information and the historical work orders of the sample data respectively to obtain binary sparse features corresponding to the worker information and the historical work orders; mapping the binary sparse features through a full connection layer of the deep collaborative filtering model to obtain a dense vector; inputting the dense vector into a deep collaborative filtering layer of the deep collaborative filtering model for vector mapping to obtain an initial predicted time; calculating a time prediction loss based on the initial predicted time and an actual work order completion time according to a preset loss function to obtain the time prediction loss; and optimizing model parameters of the deep collaborative filtering model based on the time prediction loss to obtain the target collaborative filtering model.

2. The deep collaborative filtering based ticket assignment method of claim 1, wherein, the standardization processing of the historical work order matrix corresponding to the historical data to obtain standard data comprises: performing standardization processing on the historical work order matrix according to the following formula: wherein, denotes the standard value, θ ij denotes the real completion time of worker j for order i, min denotes the minimum time required for all workers to complete order i, and max denotes the maximum time required for all workers to complete order i.

3. The deep collaborative filtering based ticket assignment method of claim 1, wherein, the inputting of the dense vector into the deep collaborative filtering layer of the deep collaborative filtering model for vector mapping to obtain an initial predicted time comprises: performing prediction on the work order completion time according to the following formula: where θ ij denotes the initial predicted time for worker j to complete order i, denotes a vector of all workers, denotes a vector of all orders, P, Q respectively denote the latent feature matrices of workers and orders, P T , Q T denote the rotation matrices of the latent feature matrices of workers and orders respectively, Θ f denote the model parameters of the multi-layer neural network f.

4. The deep collaborative filtering based ticket assignment method of claim 1, wherein, the inputting of the sample data into a preset deep collaborative filtering model for time prediction training to obtain a target collaborative filtering model further comprises: performing reverse conversion on the initial predicted time according to the following formula to map the initial predicted time back to a real value: θ true = (max - min) * θ pred + min; where θ pred represents the initial predicted time for worker j to complete order i after standardization, θ true represents the actual time for worker j to complete order i after mapping, min represents the minimum time for worker j to complete order i, and max represents the maximum time for all workers to complete order i.

5. The deep collaborative filtering based ticket assignment method of claim 1, wherein, the minimum cost calculation on the predicted time matrix corresponding to each small batch of work orders by using the Hungarian algorithm to obtain an optimal assignment matrix of each small batch of work orders comprises: the optimal assignment matrix is calculated according to the following formula: where W, K represent worker and work order set respectively, φ ij denotes the predicted time for worker j to complete work order i, z ij denotes whether to assign work order i to worker j, 1 if yes, 0 otherwise.

6. A deep collaborative filtering based ticket assignment apparatus, characterized in that, ​ The acquisition module is configured to acquire historical data of time required for workers to complete work orders, and preprocess the historical data to obtain sample data, the historical data including worker information, historical work orders, and work order completion time; The training module is configured to input the sample data into a preset deep collaborative filtering model for time prediction training, to obtain a target collaborative filtering model; The prediction module is configured to input a to-be-assigned work order into the target collaborative filtering model to predict completion time of the to-be-assigned work order for each worker, to obtain a prediction time matrix; The batch module is configured to divide the to-be-assigned work order into multiple small batches according to regions, to obtain multiple small batches of work orders; The optimization module is configured to perform minimum cost calculation on the prediction time matrix corresponding to each small batch of work orders by using a Hungarian algorithm, to obtain an optimal assignment matrix of each small batch of work orders, and perform work order assignment according to the optimal assignment matrix. The acquisition module is specifically configured to perform standardization processing on a historical work order matrix corresponding to the historical data, to obtain standard data. The standard data is subjected to anomaly detection to obtain abnormal data, the abnormal data is excluded to obtain normal data, the normal data is subjected to missing value filling and data conversion to obtain the sample data; the standard data is subjected to anomaly detection according to the following formula to obtain abnormal values: Z = (x - x m ) / s wherein Z represents an abnormal value, x represents a sample number, x m represents an average value of all samples, and s represents a standard deviation of all samples; if an absolute value of Z of a sample is greater than 3, the sample is the abnormal data; The training module is specifically configured to perform one-hot encoding on worker information and historical work orders of the sample data respectively, to obtain binary sparse features corresponding to the worker information and the historical work orders; map the binary sparse features through a full connection layer of the deep collaborative filtering model, to obtain dense vectors; input the dense vectors into a deep collaborative filtering layer of the deep collaborative filtering model for vector mapping, to obtain initial prediction time; calculate time prediction loss according to a preset loss function based on the initial prediction time and actual work order completion time, to obtain the time prediction loss; optimize model parameters of the deep collaborative filtering model based on the time prediction loss, to obtain the target collaborative filtering model.

7. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the deep collaborative filtering-based work order assignment method in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, which, when executed by a processor, causes the processor to execute the deep collaborative filtering-based work order assignment method in any one of claims 1 to 5.

Images