CN114841664A - Method and device for determining multitasking sequence - Google Patents
Method and device for determining multitasking sequence Download PDFInfo
- Publication number
- CN114841664A CN114841664A CN202210489685.5A CN202210489685A CN114841664A CN 114841664 A CN114841664 A CN 114841664A CN 202210489685 A CN202210489685 A CN 202210489685A CN 114841664 A CN114841664 A CN 114841664A
- Authority
- CN
- China
- Prior art keywords
- task
- graph
- processing
- edge
- node
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/10—Office automation; Time management
- G06Q10/103—Workflow collaboration or project management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Strategic Management (AREA)
- Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Entrepreneurship & Innovation (AREA)
- Probability & Statistics with Applications (AREA)
- Economics (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the specification discloses a method and a device for determining a multitasking sequence. The method comprises the following steps: acquiring a task relation graph; wherein, different tasks needing to determine the processing sequence are taken as different nodes in the task relation graph; the weight of each edge in the task relation graph is inversely related to the saved time corresponding to the two tasks connected with the edge; determining the weight of each edge in the task relation graph and the probability of the minimum path based on an artificial intelligence model; training an artificial intelligence model in advance according to the graph sample and the label corresponding to each edge in the graph sample, wherein the label corresponding to each edge is used for representing the weight and the minimum path of the edge which belongs to or does not belong to the graph sample; determining a path reaching the standard from all possible paths which never repeatedly pass through each node in the task relation graph according to the determined probability of each edge; and determining the processing sequence of each task according to the connection relation of each task in the standard reaching path.
Description
Technical Field
The embodiment of the specification relates to the technical field of computer application, in particular to a method and a device for determining a multitasking sequence.
Background
Currently, in many scenarios, several tasks need to be handled manually in succession. For example, in a qualification scene, a service person is usually required to continuously process a plurality of tasks for qualification verification according to information submitted by a merchant to be verified and a verification rule trained and mastered, where the tasks may specifically include determining whether the merchant to be verified belongs to legal business, determining business conditions of the merchant to be verified, and the like, so as to subsequently determine whether the merchant to be verified can pass qualification verification.
The continuous processing sequence of several tasks has a great influence on the overall processing efficiency. There is a need for a method of determining the processing order of several tasks in order to improve the overall processing efficiency.
Disclosure of Invention
In order to solve the above problems, the present specification provides a multitask order determination method and apparatus. The technical scheme is as follows.
A multitasking order determination method comprising:
acquiring a task relation graph; wherein, different tasks needing to determine the processing sequence are taken as different nodes in the task relation graph; the weight of each edge in the task relation graph is inversely related to the saved time corresponding to the two tasks connected with the edge; the calculation mode of the saved time corresponding to the first task and the second task comprises the following steps: a time length for processing only the first task in one processing, a time length for processing only the second task in another processing, and a time length for continuously processing the two tasks in one processing are subtracted;
determining the weight of each edge in the task relation graph and the probability of the minimum path based on an artificial intelligence model; training an artificial intelligence model in advance according to the graph sample and a label corresponding to each edge in the graph sample, wherein the label corresponding to each edge is used for representing the weight and the minimum path of the edge which belongs to or does not belong to the graph sample; the weight and the minimum path of the graph are the paths with the minimum weight and the minimum path in all possible paths which do not repeatedly pass through each node of the graph;
determining a path reaching the standard from all possible paths which never repeatedly pass through each node in the task relation graph according to the determined probability of each edge; the probability sum of the qualifying paths is greater than at least one other path among all possible paths that do not repeatedly pass through nodes of the task relationship graph;
and determining the processing sequence of each task according to the connection relation of each task in the standard reaching path.
A multitasking order determining apparatus comprising:
the acquisition unit is used for acquiring a task relation graph; wherein, different tasks needing to determine the processing sequence are taken as different nodes in the task relation graph; the weight of each edge in the task relation graph is inversely related to the saved time corresponding to the two tasks connected with the edge; the calculation mode of the saved time corresponding to the first task and the second task comprises the following steps: a time length for processing only the first task in one processing, a time length for processing only the second task in another processing, and a time length for continuously processing the two tasks in one processing are subtracted;
the probability unit is used for determining the weight of each edge in the task relation graph and the probability of the minimum path based on the artificial intelligence model; training an artificial intelligence model in advance according to the graph sample and a label corresponding to each edge in the graph sample, wherein the label corresponding to each edge is used for representing the weight and the minimum path of the edge which belongs to or does not belong to the graph sample; the weight and the minimum path of the graph are the paths with the minimum weight and the minimum path in all possible paths which do not repeatedly pass through each node of the graph;
the path unit is used for determining a path reaching the standard from all possible paths which never repeatedly pass through each node in the task relation graph according to the determined probability of each edge; the probability sum of the qualifying paths is at least greater than one other path among all possible paths that do not repeatedly pass through nodes of the task relationship graph;
and the sequence unit is used for determining the processing sequence of each task according to the connection relation of each task in the standard reaching path.
According to the technical scheme, the processing sequence among the multiple tasks can be determined based on the task relation graph reflecting the time saving time among the tasks, so that the overall processing efficiency of the multiple tasks is improved by increasing the total time saving time of the multiple tasks.
Drawings
In order to more clearly illustrate the embodiments of the present specification or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments described in the embodiments of the present specification, and other drawings can be obtained by those skilled in the art according to these drawings.
Fig. 1 is a flowchart illustrating a multitask processing order determining method provided in an embodiment of the present specification;
FIG. 2 is a schematic diagram illustrating a task graph construction method according to an embodiment of the present disclosure;
FIG. 3 is a flow chart of a sub-graph sampling provided by an embodiment of the present disclosure;
fig. 4 is a schematic structural diagram of a multitask processing order determining device provided in an embodiment of the present specification;
fig. 5 is a schematic structural diagram of an apparatus for configuring a method according to an embodiment of the present disclosure.
Detailed Description
In order to make those skilled in the art better understand the technical solutions in the embodiments of the present specification, the technical solutions in the embodiments of the present specification will be described in detail below with reference to the drawings in the embodiments of the present specification, and it is obvious that the described embodiments are only a part of the embodiments of the present specification, and not all the embodiments. All other embodiments derived by one of ordinary skill in the art from the embodiments given herein are intended to fall within the scope of the disclosure.
Currently, in many scenarios, several tasks need to be handled manually in succession. For example, in a qualification scene, a service person is usually required to continuously process a plurality of tasks for qualification verification according to information submitted by a merchant to be verified and a verification rule trained and mastered, where the tasks may specifically include determining whether the merchant to be verified belongs to legal business, determining business conditions of the merchant to be verified, and the like, so as to subsequently determine whether the merchant to be verified can pass qualification verification.
In a specific example, when the business platform checks the business qualification of the business, the business personnel is usually required to perform processing for related tasks according to the information submitted by the business, for example, whether the business has a license, the business range of the business, the credit evaluation of the business, and the like. The form of these tasks may include selection, statement, fill in space, and the like.
In addition, business personnel usually need to process all tasks of merchant qualification audit, so that all reasons for failing to pass the qualification audit can be comprehensively and accurately fed back to the merchant under the condition that the merchant qualification audit fails, and the success rate of the merchant for the next qualification audit is improved.
The continuous processing sequence of several tasks has a great influence on the overall processing efficiency. There is a need for a method of determining the processing order of several tasks in order to improve the overall processing efficiency.
In order to improve the overall processing efficiency of tasks, the embodiments of the present specification provide a method for determining a multitasking sequence.
In the method, considering that association relations may exist among a plurality of tasks in a plurality of scenes, when tasks with strong association are processed manually and continuously, the tasks can be processed quickly, and the overall processing efficiency is improved.
For example, in the scenario of merchant qualification audit, there are two tasks related to merchant risk assessment, namely whether the merchant has a business license and whether the merchant is operating legally.
If the business personnel determines that the merchant does not have the business license according to the merchant information, the next task is whether the merchant operates legally. Because the merchant is not legally operated under the condition that the merchant does not have a business license, business personnel can directly determine that the merchant is not legally operated according to the processing condition of the last task, and quickly process the task.
If the business personnel determine that the merchant does not have the business license according to the merchant information, dozens of tasks are processed, and then whether the merchant operates legally is processed. The business personnel may not have the related memory and need to query the merchant data again to determine that the merchant does not have the business license, and then process the task, so that the processing efficiency is low.
Obviously, when the tasks with relevance are continuously processed, a certain prompting or guiding function may exist, which is convenient for business personnel to rapidly process, improves the processing efficiency and reduces the processing time.
Therefore, the processing sequence of the tasks can be reasonably determined, so that the tasks with strong relevance can be continuously processed, and the overall processing efficiency of the tasks can be improved.
Of course, the processing sequence can be conveniently adjusted at any time according to the actual requirements and the modification of the tasks, so that the time required by the processing is saved.
The order of the tasks is specifically adjusted, on one hand, the relevance between the tasks needs to be determined, and on the other hand, an optimal processing order needs to be determined for all the tasks in consideration of the whole process, so that the processing efficiency is improved.
Therefore, in the method for determining a sequence of multitasking provided by the embodiments of the present specification, the time consumed for processing a task alone may be determined, and then the time consumed for processing the task next to another task may be determined.
Obviously, whether the task is processed independently or not saves time compared with the time consumed by continuous processing can be determined through data analysis, so that the relevance between the task and other tasks can be embodied based on the saved time.
Specifically, if the saved time is short, it may be determined that the association relationship between two consecutively processed tasks is small; if the saved time is long, the association relationship between two continuously processed tasks can be determined to be large.
Of course, the saved time itself may be used to improve the overall efficiency of task processing, and therefore, the order of task processing may be determined based on the saved time itself.
Further, a graph structure can be constructed according to tasks, the saved time is converted into weights on edges, or the weights can be converted into distances between corresponding nodes of the tasks, the distances can be inversely related to the saved time, and the longer the distance is, the shorter the saved time is. After a graph structure is constructed for all tasks needing to determine a processing sequence, the time saving relation of the whole tasks can be determined, and the incidence relation among the tasks can be represented.
For the constructed graph structure, the problem of determining the task processing order can be converted into a traveler problem, namely, the shortest path distance is ensured under the condition that each node in the graph structure is not repeatedly passed.
In other words, the shortest distance path in the graph structure may be determined, the shortest distance path may pass through each node corresponding to a task in the graph structure, the path itself may represent a task processing order, and the path distance may be negatively related to the saved duration, and in the case that the overall path distance is the shortest, the total saved duration of the corresponding task processing order is the longest, so that the duration of processing all tasks may be saved, and the overall efficiency of task processing may be improved.
Therefore, the method can construct the graph structure based on the time saved by the task continuous processing, so that the shortest distance in the graph structure can be solved according to the distance for representing the time saved in the graph structure, a better task processing sequence can be determined based on the solving result, the total time saved by the multi-task processing is increased, and the overall efficiency of the multi-task processing is improved.
Of course, under the condition that the graph structure is relatively complex, it is usually difficult to determine the global optimal solution of the shortest distance in the graph structure, and a local optimal solution, that is, a path with a relatively short total distance in the graph structure, may often be determined, so as to improve the overall efficiency of task processing.
It should be noted that, in the method embodiment provided in the embodiment of the present specification, a scene of a specific task is not limited, and multiple audit tasks for one audit object in an audit scene may be used, for example, multiple audit tasks required for auditing for a merchant. By the method, the overall processing efficiency of a plurality of audit tasks can be improved.
For example, when a certain project is evaluated, a plurality of evaluation tasks may exist by evaluating a plurality of aspects of the project. The method can improve the overall processing efficiency of a plurality of evaluation tasks.
For example, a certain questionnaire has many questions and needs to be answered, and in order to improve the overall efficiency of answering, the answering sequence of multiple questions, that is, the processing sequence of multiple answering tasks, may be determined. By the method, the overall processing efficiency of the multiple answering tasks can be improved.
The method can determine the time saved by continuous processing between tasks through data analysis, analyze the correlation property between the tasks and determine whether inspiration or correlation or association exists, thereby being capable of excavating a better multi-task processing sequence and improving the overall processing efficiency.
The following explains a multitask processing order determining method provided in the embodiments of the present specification in detail with reference to the drawings.
Fig. 1 is a schematic flowchart of a method for determining a multitasking sequence according to an embodiment of the present disclosure. The method may include the following steps.
S101: and acquiring a task relation graph.
Wherein, different tasks needing to determine the processing sequence are taken as different nodes in the task relation graph; the weight of each edge in the task relation graph is inversely related to the saved time corresponding to the two tasks connected with the edge; the calculation mode of the saved time corresponding to the first task and the second task comprises the following steps: the duration of processing only the first task in one processing, plus the duration of processing only the second task in another processing, and minus the duration of processing both tasks consecutively in one processing.
S102: and determining the weight of each edge in the task relation graph and the probability of the minimum path based on an artificial intelligence model.
Training an artificial intelligence model in advance according to the graph sample and a label corresponding to each edge in the graph sample, wherein the label corresponding to each edge is used for representing the weight and the minimum path of the edge which belongs to or does not belong to the graph sample; the weight and the minimum path of the graph are the paths with the minimum weight and the minimum path in all possible paths which do not repeatedly pass through each node of the graph.
S103: and according to the determined probability of each edge, determining a path reaching the standard from all possible paths which do not repeatedly pass through each node in the task relation graph.
The probability sum of the qualifying paths is greater than at least one other path among all possible paths that do not repeatedly pass through the nodes of the task relationship graph.
S104: and determining the processing sequence of each task according to the connection relation of each task in the standard-reaching path.
The method and the process can determine the processing sequence among the multiple tasks based on the task relation graph reflecting the time saving time among the tasks with the aim of increasing the total time saving time, thereby improving the overall processing efficiency of the multiple tasks by increasing the total time saving time of the multiple tasks.
Specifically, the overall incidence relation of the multiple tasks can be represented by introducing the saved time between the tasks based on a task relation graph reflecting the saved time between the tasks, and the path which does not repeatedly pass through each node in the task relation graph is determined by using a graph structure and an artificial intelligence model with the aim of increasing the saved total time, so that the processing sequence of the multiple tasks with relatively long saved total time can be determined, the saved total time of the multiple tasks is increased, and the overall processing efficiency of the multiple tasks is improved.
Specifically, the probability that each edge in the task relation graph belongs to the weight and the minimum path can be determined according to the task relation graph reflecting the time saving between tasks, and the weight and the minimum path can reflect the task processing sequence with the longest total time saving, so that the probability and the relatively higher standard reaching path can be determined by utilizing the graph structure and the artificial intelligence model aiming at increasing the total time saving, the task processing sequence with the relatively longer total time saving can be further determined, the total time saved by processing multiple tasks can be increased, and the overall processing efficiency is improved.
First, the following explains S101 in detail.
S101: and acquiring a task relation graph.
Wherein, different tasks needing to determine the processing sequence are taken as different nodes in the task relation graph; the weight of each edge in the task relationship graph is inversely related to the saved time corresponding to the two tasks connected by the edge.
For the saved time lengths corresponding to the two tasks, the calculation mode may be that the time length of only processing one of the tasks in one processing is added to the time length of only processing the other one of the tasks in the other processing, and the time length of continuously processing the two tasks in one processing is subtracted.
For convenience of description, the following description is made in terms of a first task and a second task.
The calculation method of the saved time corresponding to the first task and the second task may include: the duration of processing only the first task in one processing, plus the duration of processing only the second task in another processing, and minus the duration of processing both tasks consecutively in one processing.
The flow of the method is not limited to the specific manner of acquiring the task relationship diagram, and the task relationship diagram in the above form may be used.
Alternatively, a task relation graph can be constructed for several tasks for which the processing order needs to be determined.
The method flow does not limit the specific way of constructing the task relation diagram.
In an optional embodiment, a plurality of tasks may be acquired, a task having an association relationship among the plurality of tasks is determined, and a time saving duration corresponding to the association relationship is determined.
Specifically, the task relationship graph may be constructed by constructing an edge according to an association relationship between tasks for each acquired task construction node, and determining the weight of the corresponding edge according to the saved time of continuous processing of the two tasks.
The acquired tasks can be tasks needing to determine the processing sequence, and the task relation graph is convenient to construct subsequently.
1. For a plurality of tasks, the flow of the method does not limit the specific task content or form, nor the acquisition mode or source of the tasks.
For example, in a merchant admission auditing scenario, merchant information auditing tasks may be set in advance according to expert experience and actual requirements, and then several tasks may be obtained from the preset merchant information auditing tasks. The specific form of the task can be selection, filling in a blank, expression and the like, and specifically, images can be uploaded or words can be edited for processing. The specific content of the task may be to examine various information of the merchant, and specifically may include merchant operation qualification, merchant credit rating, merchant scale, merchant operation amount, and the like.
In an alternative embodiment, the task may include one or more subtasks, and may also distinguish between a core task and a derivative task, where the core task has a relatively large relationship with whether the merchant can pass the audit, for example, the legal business qualification of the merchant, and the derivative task has a relatively small relationship with whether the merchant can pass the audit, for example, the business quota of the merchant.
2. For determination of an association.
The method flow does not limit the specific method for determining the association relationship, and two examples are given below for exemplary explanation.
1) In an alternative embodiment, the association relationship may characterize the correlation between tasks, for example, in a merchant admission audit scenario, two tasks related to a merchant credit level may have a correlation, i.e., an association relationship.
Alternatively, the association relationship may be configured in advance according to the task content or the task type, for example, it is determined that there is an association relationship between all tasks related to the merchant credit level.
When the tasks with the relevance are processed continuously, the time required by the processing can be saved. For example, two tasks associated with a merchant's credit rating may be "is the merchant's credit rating determined from the user's return ratio? "and" is merchant credit rating greater than level 3? ". When the two tasks are continuously processed, the processing result of the next task can be quickly determined according to the credit level of the merchant calculated by the previous task, and the processing efficiency is improved.
Therefore, the tasks with the association relation can be continuously processed mainly by optimizing the processing sequence of the tasks in the flow of the method, so that the overall processing time is saved, and the overall processing efficiency is improved.
2) In another alternative embodiment, the association relationship may be used to characterize the continuously processed tasks, and as long as two different tasks are continuously processed, it may be determined that the two tasks have the association relationship, so as to facilitate determining whether the tasks save time in the continuous processing process from the viewpoint of data analysis.
Therefore, the tasks which are continuously processed can be determined as the tasks with the association relationship from the historical task processing data, so that the association relationship among the tasks can be conveniently mined and the time can be saved according to the historical task processing data.
And the incidence relation is analyzed from the actual processing data, the time is saved, and the non-visual incidence relation can be more accurately and directly excavated.
Combining the two embodiments, optionally, the determination of the association relationship may be specified in advance according to the task content, or may be determined according to historical task processing data analysis.
In an alternative embodiment, determining a task having an association relationship among the acquired tasks may include: acquiring historical processing data of the acquired tasks; the history processing data includes: data representing a processing sequence of a plurality of tasks among the acquired plurality of tasks; and for any task in the acquired tasks, determining a task which is processed before the task in the historical processing data, and determining that the task has an association relation with the determined task.
The embodiment does not limit the specific form of the association relationship, and optionally, the time saving duration between tasks may be determined as one form of the association relationship.
The history processing data will be explained later, and will not be described in detail here.
Because the subsequent steps need to pay attention to the saved time of the continuous processing tasks, every two continuously processed tasks can be determined to have an association relationship, and the saved time can be conveniently determined through data analysis in the subsequent steps.
It should be noted that, in the subsequent step, a graph structure needs to be constructed according to the association relationship, and a node corresponding to each task in the graph structure needs to be connected with at least one node corresponding to one other task, so optionally, each task in the obtained multiple tasks needs to have an association relationship with at least one other task.
In addition, in an alternative embodiment, the association relationship may correspond to a sequential processing order of the two tasks.
For the same two tasks, different association relationships may exist, for example, two different association relationships exist between the first task and the second task, one corresponding to the order of successively processing the first task and the second task, and the other corresponding to the order of successively processing the second task and the first task.
For example, for a first task and a second task, the historical processing data may include data characterizing that the first task is processed before the second task, and accordingly, it may be determined that the first task is associated with the second task. In addition, the association relationship between the first task and the second task may be determined, which corresponds to the order of processing the first task and the second task successively, that is, the processing order is the first task before the second task after the first task.
If the historical processing data further includes data representing that the second task is processed before the first task, it may be determined that another association relationship exists between the first task and the second task, and the another association relationship may correspond to an order in which the second task and the first task are successively processed, that is, the processing order is that the second task is previous and the first task is next.
The processing order of the two tasks corresponding to the association relationship can be specifically explained later for determining the time saving corresponding to the association relationship and establishing the influence of the task relationship existence.
3. The data is processed against history.
In an alternative embodiment, the historical processing data may include data related to a number of tasks that the business person has acquired in the past.
Optionally, the obtaining manner of the historical processing data is not limited, and specifically, the historical processing data may be obtained from a record after the record is recorded, for example, a log record, in the process of the service personnel processing.
Optionally, for the collection of the historical processing data, the relevant data may be obtained and stored by monitoring the process of processing several acquired tasks by the service personnel.
It should be noted that, when collecting historical processing data, data of service personnel in the process of processing tasks can be monitored and acquired according to the process requirement of the method. Such as processing order, processing duration, processing results, etc. The process of processing tasks may also be adapted to facilitate the acquisition of more aspects and kinds of data, e.g. adjusting the order of task processing, etc.
In an alternative embodiment, for specific content of the related data, in order to facilitate determining the association relationship between the tasks, an order in which the business person processes the acquired tasks may be included.
In general, the processing sequence of the tasks may be fixed, and the business personnel process the tasks according to the fixed sequence. However, in order to further mine the association relationship between tasks and save time, the task processing sequence can be adjusted, so that different business personnel can process tasks according to different processing sequences.
Alternatively, the order of several tasks may be randomly adjusted, processed by a service person, and the processing order recorded. The processing sequence of the tasks can be adjusted by the service personnel.
For example, for the first task-10, several service personnel may perform processing according to the processing sequence of the first task-10, and then perform processing according to the processing sequence of the first task 0-1 by other service personnel.
Therefore, optionally, data characterizing the processing order of a plurality of tasks among the acquired several tasks may be included in the historical processing data.
Of course, the data format of the specific representation processing sequence is not limited, and the representation may be performed by task identifiers having a sequence.
It should be noted that the historical processing data may include a plurality of sets of data representing different processing sequences, and each processing sequence may include all or part of the acquired tasks.
For example, the history processing data may include: first task-10, first tasks 0-7, tasks 4-9, and so on.
Obviously, through the embodiment, the association relationship between the tasks can be conveniently mined and the time can be saved by utilizing the conditions of various processing sequences between the tasks, and the association relationship between the tasks can also be helped to be determined.
In addition, in an optional embodiment, since the association relationship needs to reflect the saved time and help the subsequent steps to improve the processing efficiency, the specific content of the related data may further include the time consumed by the service personnel to process several tasks in sequence, or may be described as the processing time.
Optionally, the historical processing data may include an actual processing time length of the service personnel continuously processing the acquired plurality of tasks, and specifically may include an actual processing time length of each task in the process of continuously processing the acquired plurality of tasks.
The actual processing time may specifically refer to the time consumed by processing a task in the actual processing process of the service staff. Correspondingly, when different service personnel process the same task, the actual processing time length may be different due to various situations, and therefore, the theoretical processing time length needs to be determined according to the actual processing time length.
The theoretical processing time specifically refers to the time consumed by business personnel to process tasks theoretically, so that a fixed value can be conveniently determined for calculating the saved time.
It should be noted that, since various different processing sequences may be included in the historical processing data, the actual processing time duration for the same task to process next to different other tasks may be different.
In order to calculate the time saved by continuously processing the tasks with the association relationship, the time of individually processing the tasks needs to be determined.
In an alternative embodiment, the historical processing data may include the actual processing time duration for the service person to process each of the acquired tasks individually, i.e., the time duration for processing only one task at a time. In other words, the historical processing data may include the time duration for which the tasks are individually processed.
Obviously, through historical processing data, the actual saved time of the task being continuously processed can be conveniently determined through the obtained actual processing time of each task in the plurality of tasks being processed independently and the actual processing time of the task being processed next to other tasks.
The actual processing time length of the task to be processed next to the other task may be specifically a total actual processing time length of the two tasks processed continuously.
It should be noted that, different business persons may have different processing conditions and actual processing durations during specific processing tasks, and the historical processing data includes multiple different processing conditions of multiple business persons, for example, the same business person performs business admission check for different merchants, and the actual processing durations may also be different.
In order to facilitate the calculation of the time saving and improve the calculation accuracy, in an optional embodiment, the theoretical processing time for the task to be processed separately and the theoretical processing time for the task to be processed next to other tasks may be calculated in a summary statistical manner.
The embodiment is not limited to a specific calculation method, and may be a method of averaging, a method of finding a mode, a method of finding a median, or a method of deleting a maximum and minimum value and then averaging.
Optionally, the time consumed by the task individual processing may be determined through testing, specifically, multiple service personnel may individually process the same task for multiple times, obtain the actual processing time of the processing, and calculate the average value as the theoretical processing time.
Correspondingly, since the actual processing time may also have a deviation when different business persons continuously process two identical tasks, optionally, in order to reduce the influence of different business persons and different processing situations, the actual processing time in the historical processing data may be averaged to obtain the theoretical processing time for processing any task next to another task among the obtained multiple tasks. Of course, the theoretical processing time length for any task to process next to different other tasks may be different.
For ease of understanding, in an alternative application embodiment, the retrieved number of tasks may include a first task and a second task. After the test, the step of obtaining the actual processing time length of the first task processed by different service personnel independently comprises the following steps: 4 seconds, 5 seconds, 3 seconds, 8 seconds. Specifically, the theoretical processing time of the individual processing of the first task may be calculated to obtain a theoretical processing time of 5 seconds by calculating an average value.
The actual processing time of the first task which is continuously processed by different service personnel next to the second task comprises the following steps: 2 seconds, 3 seconds, 4 seconds, 3 seconds. Therefore, the theoretical processing time length of the first task to be processed next to the second task can be calculated to be 3 seconds by calculating an average value.
Obviously, from the perspective of conveniently calculating the time saving time, the theoretical processing time can be calculated by using the actual processing time contained in the historical processing data. The theoretical processing time period can be considered as being obtained from data analysis performed on the actual processing time period.
The step of calculating the theoretical processing time length can be executed in advance, and the theoretical processing time length can be directly used as a part of historical processing data; it can also be calculated immediately in case the saving time needs to be calculated later.
Of course, as the historical processing data can be continuously updated and increased, the situation that new service personnel process a plurality of acquired tasks is increased, and the theoretical processing duration can be updated in real time.
Optionally, the history processing data may include: acquiring actual processing time and/or theoretical processing time for independently processing any one of the plurality of tasks; the method can further comprise the actual processing time and/or the theoretical processing time of any one of the obtained tasks for processing next to other tasks.
Of course, in an alternative embodiment, the processing duration may be determined in other ways than by data analysis, for example, by expert experience and actual research analysis, and the processing duration may be set manually.
Therefore, optionally, the history processing data may include: the acquired time length for independently processing any one of the plurality of tasks; the method can also comprise the time length of any one of the acquired tasks to be processed next to other tasks. The specific determination method of the duration is not limited, and may be directly obtained actual processing duration, theoretical processing duration obtained through data analysis, or manually set processing duration, or the like.
Optionally, the acquired actual processing time length and/or theoretical processing time length of any one of the tasks to be processed next to the other tasks may be included in the historical processing data, and this data may be used to characterize the processing order of the tasks.
Therefore, the historical processing data may include data representing the processing sequence of a plurality of tasks in the acquired tasks, and the data may be in the form of actual processing time and/or theoretical processing time of any one of the acquired tasks for processing next to other tasks.
In this embodiment, the saved time can be conveniently calculated according to the historical processing data.
4. For the determination of the time saving.
The flow of the method is not limited to the determination method of the saved time.
The determination of the time saving period is explained in two ways below. On one hand, the value of the saved time length is taken, and on the other hand, the two tasks corresponding to the saved time length are processed sequentially.
1) Saving the value of time length.
Optionally, the saved time between tasks may be determined by historical processing data and by a data analysis manner, or may be determined by manual setting.
In an alternative embodiment, the historical processing data may include the duration of the acquired tasks that are processed separately, that is, the duration of processing only one task at a time; the method may further include continuously processing the duration of the two tasks in the obtained plurality of tasks, specifically, continuously processing the duration of the two tasks in one of a plurality of combinations including the two tasks.
Correspondingly, the saving time of the continuous processing can be determined by simple difference calculation aiming at any combination.
Taking the first task and the second task as an example, the calculation method of the saved time corresponding to the first task and the second task may include: the duration of processing only the first task in one processing, plus the duration of processing only the second task in another processing, and minus the duration of processing both tasks consecutively in one processing.
Of course, the time duration therein may be a calculated theoretical processing time duration.
Optionally, the actual saved time may be calculated according to the actual processing time of the task, and then the theoretical saved time may be further calculated as the saved time corresponding to the association relationship.
The actual saved time specifically may refer to actual time saved in an actual processing process of service personnel, and actual time saved by different service personnel may be different, so that actual saved time of the same task may be integrated to determine a theoretical saved time. The theoretical time saving specifically may refer to the theoretical time saving of business personnel in continuous task processing.
Optionally, the theoretical processing duration may be calculated first and then the theoretical saved duration may be calculated according to the actual processing duration of the task. Or the theoretical saved time can be directly calculated according to the theoretical processing time.
In general, when tasks are processed continuously, if there is no correlation between the task processed before and the task processed currently, and there is no hint, guidance or prompt, then the two tasks can be regarded as being processed separately.
Thus, the calculated savings time is typically a positive number or zero.
It is noted that the calculated difference may be zero, since the determination of the correlation may be directly from the data analysis, which merely characterizes the continuously processed tasks, which may not be relevant and may not actually save time.
However, when data is analyzed, the theoretical processing time may be inaccurate, so that the theoretical processing time of the single processing is shorter than the theoretical processing time of the continuous processing, and further, the calculated difference may be negative. For example, the actual processing time length of the individual processing of a certain task included in the historical processing data is small in number, so that the theoretical processing time length of the individual processing is large in error, and a negative number may occur in the subsequent calculation of the difference.
The reason for calculating the difference to be negative is due to errors in the data analysis process. In order to avoid the influence caused by the error, the saving time is generally considered to be 0, that is, the continuous processing does not save the processing time.
In addition, in some special cases, there may be misleading or confusing cases in actual processing, such as a task "whether the merchant has an entity store" and a task "business entity name of the merchant". For business personnel who are not familiar with relevant knowledge, the physical store and the business entity may be confused, so that misleading situation occurs, and further, the continuous processing of the two tasks can increase the processing time, even longer than the processing time of the single processing task.
It can be seen that, for the case where the calculated difference is negative, it may be an error in the data analysis process, or it may be an actual case, that is, the continuous processing is delayed to reduce the efficiency.
Therefore, for the case that the difference is a negative number, when the time saving duration corresponding to the association relationship is determined, the case can be regarded as a real case, and the difference is directly determined as the time saving duration corresponding to the association relationship; it can also be considered that both are errors, and in the case where the difference is negative, the saving time period is determined to be 0.
Of course, alternatively, the error and the real situation can be considered comprehensively, and the saved time length can be determined separately. Specifically, it may be determined that the saving time period is 0 directly according to the difference threshold, if the calculated difference is a negative number and the calculated difference is greater than the preset difference threshold, the difference may be regarded as an error.
And if the calculated difference is a negative number and the calculated difference is less than or equal to the preset difference threshold, the situation can be considered as a real situation, and the calculated difference is determined as the time saving duration.
The embodiment is not particularly limited to the determination method for saving the time length when the calculated difference is negative.
Therefore, optionally, for the difference between the duration of the individual processing and the duration of the continuous processing, the difference may be directly determined as the saving duration, or the saving duration may be determined as 0 according to the difference smaller than 0.
In other words, optionally, the calculation manner of the saved time corresponding to the first task and the second task may further include: the duration of processing only the first task in one processing, plus the duration of processing only the second task in another processing, and minus the duration of processing both tasks consecutively in one processing.
When the obtained calculation result is greater than or equal to 0, the calculation result can be directly determined as the saving time length corresponding to the two tasks.
In the case that the obtained calculation result is less than 0, it may be determined that the saving time duration corresponding to the two tasks is 0.
Obviously, by the above embodiment, the value of the saved duration can be determined based on the historical processing data.
2) And the two tasks corresponding to the time length are processed in sequence.
In an alternative embodiment, the association relationship may correspond to a sequential processing order of the two tasks. Two corresponding incidence relations with different sequential processing sequences may exist between the two tasks.
In other words, the sequential order of the successive processing may be different for the two tasks of the successive processing.
Accordingly, two tasks are processed consecutively, but in a different order, and the time saved may be different, e.g., the processing of a first task implies the processing of a second task, while the processing of the second task implies no processing of the first task. Therefore, a certain time can be saved by processing the first task and the second task successively, and the time can not be saved by processing the second task and the first task successively.
Thus, in an alternative embodiment, the time savings may also correspond to a sequential processing order of the two tasks. For example, the saved duration of continuous processing for a first task to be processed next to a second task may be different from the saved duration of continuous processing for a second task to be processed next to the first task.
For two tasks, under the condition that only one saved time length is determined, the saved time length can be directly utilized to carry out subsequent steps.
And under the condition that two time saving lengths corresponding to different sequential processing sequences are determined, the two time saving lengths can be used for subsequently constructing a task relation graph, specifically a directed graph can be constructed, and the two time saving lengths can be integrated to determine one time saving length for subsequently constructing an undirected graph.
The embodiment does not limit the specific integration method, and alternatively, may be calculating an average value.
Therefore, optionally, in a case that the task relationship graph is a directed graph, the weight of each edge in the task relationship graph is inversely related to the saved time corresponding to the two tasks connected by the edge.
Under the condition that the directed edge points to the second task from the first task, the corresponding calculation mode of saving time length may include: the duration of processing only the first task in one processing, plus the duration of processing only the second task in another processing, minus the duration of processing the first task and the second task consecutively in succession in one processing.
Under the condition that the directed edge points to the first task from the second task, the corresponding calculation mode of saving time length may include: the duration of processing only the first task in one processing, plus the duration of processing only the second task in another processing, minus the duration of processing the second task and the first task consecutively in succession in one processing.
Optionally, when the task relationship graph is an undirected graph, the weight of each edge in the task relationship graph is inversely related to the saved time corresponding to the two tasks connected to the edge.
For the first task and the second task connected to the edge, the corresponding calculation mode for saving time length may include: the duration of processing only the first task in one processing, plus the duration of processing only the second task in another processing, and minus the duration of processing the first task and the second task consecutively in one processing.
In the case where only the time length for continuously processing the second task and the first task in sequence in one processing can be obtained, the time length for continuously processing the first task and the second task may be the obtained time length.
In the case where only the time length for continuously processing the first task and the second task in sequence in one processing can be obtained, the time length for continuously processing the first task and the second task may be the obtained time length.
Specifically, the duration for continuously processing the first task and the second task may be an average of two obtained durations under the condition that the durations for continuously and successively processing the second task and the first task in one processing and the durations for continuously and successively processing the first task and the second task in one processing can be obtained.
Of course, the two acquired durations may be calculated by other calculation methods. The present embodiment is not limited.
In an alternative embodiment, time savings may be aided by building a matrix.
Wherein the retrieved number of tasks may include N tasks, q1-qN respectively.
A time saving duration matrix a of N x N may then be constructed. In the saved duration matrix a, Aij may characterize the duration saved by the processing of the task qi next to qj. Wherein, i ═ 1, 2, 3.., N; j ═ 1, 2, 3.
Of course, Aii may be set to 0 and not considered later, since there is no case where qi is processed next to qi.
By historical processing of the data, a saving duration matrix a can be calculated, wherein some elements may not be calculated due to the absence of records or data, and can be temporarily defaulted.
Then, in order to calculate the average value and obtain the saving time corresponding to the association relationship, the saving time matrix a may be symmetric, specifically, Aji may be made equal to Aij by averaging.
Further, if Aij is default and Aji has a saved time length, it means that there is only a case where qj is processed after qi in the history processing data, and therefore, the value of Aji may be directly assigned to the default Aij so that Aji becomes Aij.
In this embodiment, the calculation may be aided by a matrix tool. The step of determining the weights of the edges based on the saved time period may be performed using a matrix tool.
In the embodiment of the method, the incidence relation is determined and the time is saved for the obtained tasks, so that the subsequent construction of a task relation graph is facilitated for analysis, a better task processing sequence is determined, and the overall task processing efficiency is improved.
The saved time of the continuous processing tasks can be determined by analyzing the historical processing data, and then a task relation graph can be constructed and analyzed based on the saved time.
It should be noted that since the tasks themselves may change, the historical processing data may also be accumulated and updated continuously. The association and duration savings may also vary.
For example, as the proficiency of business personnel increases, the individual processing time for certain types of tasks may be reduced, and correspondingly, the time saved may be reduced; it is also possible that as the scene changes, the task content is updated, and the retrieved tasks are also updated.
Therefore, the method embodiment can be executed periodically or aperiodically, and the association relationship and the time saving are determined by using the current latest tasks and historical processing data, so that the task relationship graph is convenient to update, and the better task processing sequence is further determined.
Optionally, several current latest tasks which need to determine the processing order or update the processing order and current latest historical processing data may be obtained in real time, and the above method embodiment is executed to determine the association relationship between the tasks and save time, so as to facilitate the subsequent construction of the task relationship graph.
5. And constructing a task relation graph based on the plurality of tasks and the determined incidence relation.
In an alternative embodiment, constructing the task relationship graph based on the obtained tasks and the determined association relationship may include: constructing a corresponding node for each task in the obtained tasks; and for each determined incidence relation, determining nodes corresponding to two tasks contained in the incidence relation, and generating an edge corresponding to the aimed incidence relation between the two determined nodes.
Because the processing sequence of the acquired tasks needs to be determined according to the task relationship diagram, and the processing sequence needs to include each acquired task, the task relationship diagram needs to include a node corresponding to each acquired task.
In an alternative embodiment, the association relationship may correspond to a sequential processing order of the tasks, and therefore, when the corresponding edge is generated according to the association relationship, a directed edge or an undirected edge may be generated.
Specifically, the directional edge is generated, and the direction of the edge may be determined according to the sequence of the continuous processing of the tasks corresponding to the association relationship.
For example, a directed edge may be generated that points from a first task to a second task for an incidence corresponding to successively processing the first task and the second task.
In addition, it should be noted that, for the constructed task relationship diagram, since a better task processing order needs to be determined by calculation subsequently, the task relationship diagram needs to include at least two paths that do not repeatedly pass through each node in the task relationship diagram, so that the better path can be selected and determined to determine the better task processing order. Otherwise there is no choice.
Optionally, in the history processing data, for a plurality of tasks whose processing order needs to be determined, different orders of at least two consecutive processing of all tasks may be included, so that the constructed task relationship graph may include at least two paths that pass through each node in the task relationship graph without repetition according to the determined association relationship.
For example, for 10 tasks from the first task to the tenth task, the historical processing data may include a processing order from the first task to the tenth task and a processing order from the tenth task to the first task, so that the constructed directed graph, that is, the task relationship graph, may include at least a path that passes through the first task to the tenth task without repetition and a path that passes through the tenth task to the first task without repetition.
Optionally, edges may be further added to the constructed task relationship graph without a corresponding association relationship, so that the task relationship graph includes at least two paths that do not repeatedly pass through each node in the task relationship graph. In particular, a fully connected graph may be constructed.
Of course, optionally, since there is no corresponding association relationship between the newly added edges and there is no corresponding time saving duration, the corresponding time saving duration may be directly set to 0.
In order to facilitate understanding of the above method embodiment for constructing a task relationship diagram, a specific application embodiment is provided below. Fig. 2 is a schematic diagram illustrating a task relationship graph building method provided in an embodiment of the present disclosure.
Including the saving duration matrix B.
B is a 3 x 3 matrix comprising 3 tasks, q1-q 3.
Bij may represent the time savings for qi to process next in qj.
Wherein, B11, B22 and B33 can be directly set to zero. B12 is 3, B21 is 4, B13 is 2, B31 is 2, B23 is 3, B32 is 4.
Thereafter, a task relationship graph may be constructed based on the matrix B.
Wherein 3 nodes are constructed for q1-q3, respectively, nodes 1-3.
Thereafter, edges may be constructed from the associations in matrix B.
According to B12, an edge may be constructed that points from node 2, corresponding to q2, to node 1, corresponding to q 1.
According to B21, an edge may be constructed that points from node 1, corresponding to q1, to node 2, corresponding to q 2.
According to B13, an edge may be constructed that points from node 3, corresponding to q3, to node 1, corresponding to q 1.
According to B31, an edge may be constructed that points from node 1, corresponding to q1, to node 3, corresponding to q 3.
According to B23, an edge may be constructed that points from node 3, corresponding to q3, to node 2, corresponding to q 2.
According to B32, an edge may be constructed that points from node 2, corresponding to q2, to node 3, corresponding to q 3.
In the subsequent step, the weight may be determined based on the saved duration corresponding to the association relationship.
6. And determining the weight of the corresponding edge based on the saved time corresponding to the incidence relation, wherein the determined weight is negatively related to the saved time.
In other words, for each edge in the task relationship graph, the weight of the edge may be determined according to the saved time lengths corresponding to the two tasks connected by the edge. The determined weight is inversely related to the saving time duration.
The embodiment does not limit the specific method for determining the weight, as long as the weight of the edge is inversely related to the saved time corresponding to the two tasks connected with the edge. The longer the saving time, the smaller the weight.
In an alternative embodiment, the weight of the edge in the task relationship graph may be regarded as the distance between the nodes, the smaller the distance is, the longer the time saving is, and for a path that does not repeatedly pass through all the nodes, the smaller the weight sum of the included edges is, the longer the total time saving of the task processing sequence corresponding to the path is.
Several alternative embodiments are provided below for the method of determining the weights of the edges for illustrative purposes.
In the first embodiment, the maximum saved time length may be determined according to the saved time lengths corresponding to two tasks connected to each edge in the task relationship graph, the standard saved time length greater than or equal to the maximum saved time length is determined, and the difference between the standard saved time length and the saved time length corresponding to each association relationship is determined as the weight of the edge corresponding to the association relationship.
Specifically, the standard saved time length is used to subtract the saved time length corresponding to each association, and the obtained difference is determined as the weight of the edge corresponding to the association.
In this embodiment, by determining the standard saving time, the distance between the saving time corresponding to each association and the standard saving time can be accurately and intuitively determined, and the distance between the nodes, that is, the weight of the edge, is conveniently determined.
In a second embodiment, the saved time lengths corresponding to the association relations corresponding to all the edges in the task relation graph may be normalized and all the saved time lengths are mapped into a range of 0 to 1, and then the difference between 1 and the normalized result of the saved time length corresponding to each association relation may be determined as the weight of the edge corresponding to the association relation.
According to the embodiment, the saved time can be converted into a value in a range of 0-1 through normalization, so that subsequent model processing is facilitated.
Second, the following explains S102.
S102: and determining the weight and the probability of the minimum path of each edge in the task relation graph based on the artificial intelligence model.
Training an artificial intelligence model in advance according to the graph sample and a label corresponding to each edge in the graph sample, wherein the label corresponding to each edge is used for representing the weight and the minimum path of the edge which belongs to or does not belong to the graph sample; the weight and the minimum path of the graph are the paths with the minimum weight and the minimum path in all possible paths which do not repeatedly pass through each node of the graph.
The method is used for determining a better processing sequence aiming at a plurality of tasks needing to determine the processing sequence and improving the overall processing efficiency.
One way to increase the processing efficiency is to reduce the time to process all tasks, in other words, to increase the total time savings as much as possible.
On the basis of the task relationship graph, a path which does not repeatedly pass through all the nodes can be determined, and the path can correspond to the processing sequence among all the tasks because all the tasks have corresponding nodes in the task relationship graph. Specifically, the order in which the path passes through the nodes may be determined as the processing order of the tasks corresponding to the nodes.
For example, for a task relationship graph including nodes 1 to 3, the nodes 1 to 3 correspond to the first task to the third task, respectively, and for a path "node 1-node 3-node 2", it may be determined that the processing order from the first task to the third task is "first task-third task-second task" according to the correspondence between the nodes and the tasks.
Therefore, on the basis of the task relationship diagram, the processing sequence of all tasks, that is, the processing sequence of several tasks requiring determination of the processing sequence, can be represented by the task relationship diagram without repeating the path passing through all nodes.
Further, since the weight of the edge is inversely related to the saving time in the task relationship graph, the smaller the weight is, the longer the saving time is.
Optionally, since a path that does not repeatedly pass through all nodes in the task relationship graph may represent a processing order of all tasks, and the path includes edges between the nodes, a sum of weights of the edges included in the path may represent a sum of saved durations of the processing order of the tasks represented by the path. The sum of the weights is inversely related to the sum of the saving durations, i.e. the smaller the sum of the weights, the longer the total saving duration.
Therefore, optionally, a path that does not repeatedly pass through all nodes in the task relationship graph may represent the processing order of all tasks; the sum of the weights of the edges included in the path can represent the sum of the time length saved under the condition that all the tasks are processed according to the processing sequence represented by the path.
Therefore, the problem of determining a better processing order among several tasks can be converted into the problem of determining weights and smaller paths without repeatedly passing through all nodes from the task relationship graph.
Of course, alternatively, in order to determine the optimal processing sequence of several tasks, so that the total saving time of the optimal processing sequence is the longest, the optimal path may be determined on the basis of the task relation graph. The optimal path may pass through each node in the task relationship graph without repetition, so that the processing order of all tasks may be characterized and the total time saved by the characterization is made longest.
And the sum of the weights of the edges included in the path in the task relationship graph can represent the total saved time, so that optionally, when the optimal path is determined, a path with the longest total saved time in the task relationship graph, that is, a path with the smallest sum of the weights of the edges included in the task relationship graph, can be determined.
In an alternative embodiment, in the task relationship graph, the weight of the edge may be regarded as the distance between two connected nodes, so that the task relationship graph may be regarded as a node distance graph. The smaller the distance between nodes (the smaller the weight of the edge between nodes), the longer the saving time.
In other words, since it is necessary to determine the weight and the smaller path without repeatedly passing through all the nodes from the task relationship diagram, and the weight of the edge can be regarded as the distance between two nodes connected by the edge, the problem of determining the better processing order among several tasks can be converted into the problem of determining the path with the shorter sum of the distances that has not repeatedly passed through all the nodes from the task relationship diagram.
For convenience of description, the path required to be determined in the process of the method is referred to as a reach path.
Alternatively, the qualifying path may be a path with a relatively small sum of weights (sum of node distances) among all possible paths that do not repeatedly traverse the nodes of the task relationship graph.
Of course, alternatively, the qualifying path may be the one with the smallest sum of weights (sum of node distances) among all possible paths that do not repeatedly pass through the respective nodes of the task relationship graph.
The embodiment does not limit the method for determining the standard reaching path.
Alternatively, all possible paths that pass through each node of the task relationship graph without repetition may be traversed, and the corresponding weight sum may be determined, so that the path with the smallest weight sum may be determined as the path reaching the standard. In this embodiment, the determined standard-reaching path has the smallest weight sum, and the corresponding total saved time is longest, so that the task processing sequence with the longest total saved time can be represented, and the overall processing efficiency of the tasks is improved.
In other embodiments, for complex task relationship graphs, it is often difficult to traverse all possible paths that do not repeatedly pass through the nodes of the task relationship graph.
For example, it is first necessary to traverse each node in the task relationship graph as a starting node to determine whether there are all possible paths from the starting node that do not repeatedly pass through the nodes of the task relationship graph. When the number of nodes in the task relationship graph is large, traversal is often difficult.
Therefore, the specific determination of the path to reach the standard may be a local optimal solution determined when solving the weight and the minimum path in the task relationship graph.
Thus, optionally, the weighted sum of the qualifying paths is greater than at least one other path in all possible paths that do not repeatedly traverse the respective nodes of the task relationship graph.
And solving the weight and the minimum path in the task relational graph, wherein the weight in the task relational graph can be regarded as the distance between the nodes, so the solution can be carried out by a method for solving the path with the minimum distance between the nodes in the graph structure.
The determination can be specifically carried out by a solution method of the traveler problem.
In an alternative embodiment, there are currently a variety of methods for solving the traveler problem, and methods for solving the exact solution of the traveler problem, such as graph neural network models, graph convolution neural network models, graph attention network models, concode, and the like.
In an alternative embodiment, weights and minimum paths in the task relationship graph may be solved using a pre-trained artificial intelligence model.
Optionally, for the artificial intelligence model, the artificial intelligence model may be trained in advance according to the graph sample and the label corresponding to each edge in the graph sample, where the label corresponding to each edge is used to represent the weight and the minimum path of the edge that may or may not belong to the graph sample.
The weight and the minimum path of the graph may be the path with the minimum weight and the minimum path among all possible paths that do not repeatedly pass through the nodes of the graph. The weight may be a weight of an edge, and may specifically characterize a distance between nodes.
In the embodiment, by training in advance, the artificial intelligence model can output the probability that each edge belongs to the weight and the minimum path in the graph structure for the input graph structure.
The embodiment does not limit the specific form of the artificial intelligence model, as long as the probability that each edge belongs to the weight and the minimum path in the graph structure can be output for the input graph structure. Alternatively, the artificial intelligence model may be a graph convolution neural network model.
The atlas neural network model may include an input layer, an atlas layer, a fully connected layer, and a softmax layer.
Specifically, a graph structure may be input into a graph convolution neural network model, a characterization vector (embedding) of each edge in the graph structure is obtained through a graph convolution layer, then, a probability of belonging to a weight and a minimum path in the graph structure may be predicted for each edge through a full connection layer, and model training may be performed with a difference between a probability of belonging to the weight and the minimum path in the graph structure (i.e., an optimal solution of a traveler problem) of each edge output by a softmax layer and a case of an actual weight and a minimum path in the graph structure as a loss.
In this embodiment, the weight and the probability of the minimum path that each edge in the task relationship graph belongs to the task relationship graph can be determined through a pre-trained artificial intelligence model, so that the standard-reaching path can be conveniently determined subsequently, and specifically, the local optimal solution can be determined.
Optionally, the task relationship graph may be input into the artificial intelligence model, and the weight and the probability of the minimum path of each edge in the task relationship graph may be output.
Correspondingly, the qualifying path may be determined according to the probability of each edge of the output.
In the embodiment, the artificial intelligence model can be directly utilized to carry out probability prediction on the task relation graph, and the calculation efficiency is improved.
It should be noted that, because the total number of tasks that need to determine the processing order is not fixed, and the tasks and the corresponding task relationship graphs may also be updated based on different situations, the artificial intelligence model often needs to predict the probability for the task relationship graphs with different numbers of nodes.
In an optional embodiment, in order to reduce the amount of computation and adapt to a variety of task quantity scales, a method of obtaining a sub-graph by splitting a task relationship graph may be used, a plurality of sub-graphs are split for the task relationship graph, and then a pre-trained artificial intelligence model may be used for each sub-graph to output the weight and the probability of the minimum path of each edge in the sub-graph belonging to the sub-graph.
Then, the probability corresponding to any edge in each sub-graph can be synthesized, and the weight of the edge belonging to the task relation graph and the probability of the minimum path are determined.
The task relationship graph is split into a plurality of sub-graphs, the complexity of the sub-graphs is usually smaller than that of the task relationship graph, and particularly, the number of nodes or edges contained in the sub-graphs can be smaller than that of the task relationship graph, so that probability prediction is performed on each sub-graph by using a pre-trained artificial intelligence model, calculation can be performed conveniently and rapidly, and calculation efficiency is improved.
Optionally, probability prediction can be performed on each sub-graph in parallel by using a pre-trained artificial intelligence model, so that fast calculation is facilitated, and calculation efficiency is improved.
Aiming at the task relational graph with different node quantity scales, the calculation efficiency can be improved in a mode of obtaining the subgraph by splitting, the method is suitable for tasks with different quantities, and the method for solving the optimal path probability set does not need to be replaced or updated according to the quantity of the tasks.
Even if the number of tasks needing to determine the processing sequence is changed, the number of the tasks is changed in a mode of obtaining subgraphs through splitting.
For example, for the task of merchant admission and audit, the task may be updated along with the change of the expert experience and the actual situation, and specifically, the situation that the number of tasks changes may exist.
Therefore, optionally, determining, based on the artificial intelligence model, a weight and a probability of each edge in the task relationship graph belonging to the task relationship graph, may include: splitting a plurality of subgraphs aiming at the task relation graph; each node in the task relation graph at least belongs to one subgraph; inputting each sub-graph into an artificial intelligence model, and outputting the weight and the minimum path probability of each edge in the sub-graph, wherein each edge belongs to the sub-graph; and aiming at each edge in the task relationship graph, acquiring the weight and the minimum path probability of the edge in each sub-graph, belonging to the sub-graph, and integrating all the acquired probabilities to determine the weight and the minimum path probability of the edge belonging to the task relationship graph.
The above steps are explained in detail below.
1. And splitting the task relation graph to obtain a subgraph.
The embodiment does not limit the method for obtaining the subgraph by specific splitting, and since the probabilities predicted by the artificial intelligence model for each subgraph need to be synthesized subsequently, it can be optionally limited that each node in the task relationship graph is divided into at least one subgraph, or each edge in the task relationship graph is divided into at least one subgraph.
Optionally, different subgraphs obtained by splitting may include the same node, or may not include the same node. Different subgraphs may or may not contain the same edge. The present embodiment is not particularly limited.
For ease of understanding, several alternative embodiments are given below to explain the method of splitting subgraphs, and these embodiments are for illustrative purposes only.
In a first embodiment, the following steps may be performed in a loop until each node in the task relationship graph is divided into at least one subgraph: selecting nodes which are not divided into any subgraph currently in the task relationship graph as initial nodes, wherein the initial nodes can be used for constructing the subgraph; then, based on the starting node, a plurality of other nodes in the task relationship graph, which are directly or indirectly connected with the starting node, are divided into a new subgraph, and edges connecting the starting node and the other nodes are divided into the new subgraph, so that a new subgraph can be obtained.
In a second embodiment, the following steps may be performed in a loop until each edge in the task relationship graph is divided into at least one sub-graph: selecting an edge which is not divided into any subgraph currently in the task relationship graph, and taking two nodes connected with the edge as initial nodes which can be used for constructing the subgraph; then, based on the starting node, a plurality of other nodes in the task relationship graph, which are directly or indirectly connected with the starting node, are divided into a new subgraph, and edges connecting two starting nodes and edges connecting the starting node and the other nodes are divided into the new subgraph, so that a new subgraph can be obtained.
For the two embodiments, optionally, when other nodes are selected to be divided into new subgraphs, the sum of the weights of the paths may be determined according to the paths connecting the other nodes and the starting node, and other nodes with corresponding weights and smaller than the preset weight value are selected, or other nodes with the first few sequence numbers in the sequencing result are selected according to the corresponding sum of the weights and the sequence from small to large. Specifically, other nodes with sequence numbers smaller than the preset sequence number in the sorting result may be selected.
In this embodiment, since the weight and the minimum path in the task relationship graph need to be solved, the weight and the relatively smaller path may be selected to construct the sub-graph, so as to add the edge included in the optimal path in the task relationship graph to the sub-graph as much as possible.
In addition, in an optional embodiment, in order to further improve the computational efficiency, it may be defined that, in the sub-graph obtained by splitting, the number of nodes included in different sub-graphs is the same.
Optionally, when the probability is predicted for the sub-graphs by using the pre-trained artificial intelligence model, because the number of nodes contained in different sub-graphs is the same, the input form of the artificial intelligence model can be fixed, thereby facilitating the deployment and training of the model.
Therefore, optionally, when the artificial intelligence model is trained in advance, the number of nodes in the pattern book for training may be limited, so that the number of nodes in all the pattern books for training the artificial intelligence model is the same. The input of the trained artificial intelligence model may also include a graph structure with a fixed number of nodes, which may be specifically the same as the number of nodes of the graph.
Correspondingly, splitting a plurality of subgraphs for the task relationship graph may include: splitting a plurality of sub-graphs containing a preset number of nodes aiming at the task relation graph; the preset node number is the node number of the graph sample.
Therefore, in order to facilitate defining the number of split sub-graph nodes, optionally, splitting a plurality of sub-graphs specifically for the task relationship graph may include: aiming at the task relationship graph, the following steps are executed in a circulating mode until each node in the task relationship graph at least belongs to one sub graph: selecting a node which does not belong to any subgraph in the task relationship graph, and determining the node as a central node; determining other nodes meeting preset screening conditions as alternative nodes in other nodes except the central node; for each alternative node, determining the path weight sum from the central node to the alternative node; according to the determined weight sum, sequencing the alternative nodes in a sequence from small to large; determining the alternative nodes with the sequence numbers smaller than the preset sequencing sequence number in the sequencing result as the sub-graph nodes; and forming a subgraph by the central node, the determined subgraph nodes and the path from the central node to each subgraph node.
The other nodes may be any node other than the central node among all nodes included in the task relationship graph. In order to improve the calculation efficiency and save the calculation resources, part of other nodes in the task relation graph can be screened out according to the preset screening conditions to be determined as alternative nodes, and then the weight sum is calculated.
Of course, the path from the center node to each of the other nodes may be directly determined without screening, and the sum of the weights of all the edges included in the determined path may be calculated.
The preset screening condition is not specifically limited in this embodiment, and optionally, the preset screening condition may be used to screen out other nodes closer to the central node in the task relationship graph. The method specifically comprises the following steps: the number of edges included in the path from the center node to the other nodes is less than the preset number of edges.
It should be noted that, in the case that the subgraph needs to include a fixed number of nodes, the number of other nodes screened by the preset screening condition needs to be at least greater than or equal to the fixed number-1. Specifically, the number of other nodes screened by the preset screening condition may be limited to be greater than or equal to the preset sorting order number. The pre-set ordering order number may be used to define the number of nodes in the subgraph.
Alternatively, the determined path from the central node to each of the candidate nodes may be a loop-free path, such that the determined path is defined not to repeatedly traverse the same node.
Optionally, the preset sorting order number may be the same for each sub-graph, so that multiple sub-graphs with the same number of nodes may be split. Of course, the preset sort order number may also be different for each sub-graph.
Optionally, the preset sorting order number may be preset, and specifically may be determined according to a preset fixed number of the sub-graph nodes. In this embodiment, by presetting the sorting sequence number, the number of nodes constituting the sub-graph can be limited, so that the number of nodes in the sub-graph can be controlled to be a fixed value.
Optionally, the loop stopping condition in this embodiment may also be that until each edge in the task relationship graph belongs to at least one sub-graph.
For ease of understanding, a specific application example is provided below.
Fig. 3 is a schematic flow chart of sub-graph sampling provided in an embodiment of the present disclosure.
The task relationship graph can include n nodes, sub-graph sampling needs to be performed on the task relationship graph to obtain a sub-graph set, and each sub-graph in the sub-graph set includes m nodes. n may specifically be 100 and m may specifically be 16.
S301: and initializing the sampling times of all nodes in the task relation graph to be 0.
S302: and randomly selecting a node with the minimum sampling times from the task relation graph as an initial node.
S303: and taking the initial node as a clustering center, selecting 15 other nodes with the shortest association distance with the initial node in the task relation graph, and connecting the initial node and the selected other nodes to form a subgraph, adding the subgraph to a subgraph set, and increasing the sampling times of all nodes in the subgraph by 1.
The association distance may specifically be a sum of weights of edges included in a path from the initial node to the other nodes.
S304: and judging whether all the nodes in the task relation graph are sampled, namely judging whether the sampling times of all the nodes in the task relation graph are more than 0. If yes, the flow ends, and if no, S302 is executed.
By the application embodiment, the subgraphs with the same number of the nodes can be obtained based on the task relationship graph.
In the embodiment of splitting the task relationship graph to obtain the subgraph, the subgraphs with the same number of nodes can be obtained to adapt to different task scales.
For example, a task relation graph including 100 nodes can be constructed for 100 acquired tasks, and then the task relation graph can be split into a plurality of subgraphs including 16 nodes, and then probability prediction can be performed for each subgraph including 16 nodes by using a pre-trained artificial intelligence model.
Even for 1000 acquired tasks, a task relation graph containing 1000 nodes can be constructed, further the task relation graph can be split into a plurality of subgraphs containing 16 nodes, and then probability prediction can be carried out on each subgraph containing 16 nodes by using the same artificial intelligence model. The number of split subgraphs may be different.
Therefore, for different task scales, subgraphs with the same number of nodes can be obtained by splitting the task relational graph, the same solving method is used, the solving method does not need to be adjusted according to the prediction probability of each subgraph, the same model can be used repeatedly, the influence of the number of tasks is avoided, and the method is suitable for different task scales.
2. And inputting each sub-graph into an artificial intelligence model, and outputting the weight and the minimum path probability of each edge in the sub-graph belonging to the sub-graph.
In an optional embodiment, because the complexity of the sub-graph is less than that of the task relation graph, the probability prediction is performed on the sub-graph by using the pre-trained artificial intelligence model, so that the calculation efficiency can be improved. Moreover, probability prediction can be performed on different subgraphs in parallel, and the calculation efficiency is further improved.
In an alternative embodiment, the number of nodes of the obtained sub-graph may be a fixed value, so that training of the artificial intelligence model may be facilitated.
Alternatively, since the number of nodes of the graph structure (sub-graph) of the artificial intelligence model needs to be input is a fixed value, the input of the model may be the graph structure of the fixed number of nodes when constructing the artificial intelligence model. For example, a graph structure (sub-graph) that requires model input may be a graph structure that contains 16 nodes, and thus, when building an artificial intelligence model, the model input may be set to a graph structure that contains 16 nodes.
In particular, the graph structure as an input of the model may be represented by a matrix, in particular by a matrix C of 16 × 16, and any element in the matrix may characterize an edge between nodes and/or a weight of the edge. For example, Cij may represent the weight of an edge between node i and node j.
Of course, since there is usually no edge connecting a node to itself in the graph structure, Cii can usually take 0 or null, indicating that there is no edge.
Similarly, if there is no edge between node i and node j in the graph structure, Cij may also be 0 or null, which indicates that there is no edge.
Thus, optionally, the input to the artificial intelligence model may be set to a matrix of 16 x 16.
When the artificial intelligence model is trained by using the pattern book, the number of graph structure nodes of the input model is a fixed value, so that a graph sample can be conveniently constructed, and a graph structure with a plurality of fixed node numbers is constructed for training.
Moreover, because the number of the nodes is usually small, the constructed pattern book can be conveniently marked, and the weight and the minimum path of the graph sample can be conveniently and quickly determined to be used as the label of the graph sample.
Therefore, in the embodiment that the number of nodes of different subgraphs is the same, training of the artificial intelligence model can be facilitated, on one hand, a large number of labeled training samples can be constructed quickly, on the other hand, the model can be constructed conveniently, and the model training efficiency is improved.
The output of the artificial intelligence model, in turn, can generally include the probability that each edge in the subgraph belongs to the weight and minimum path in the subgraph.
Optionally, the model output may be in the form of a matrix, and specifically, for a graph structure including m nodes, the model output may be in the form of a matrix D of m × m. Dij may represent the probability that the edge between node i and node j belongs to the weight and the smallest path in the graph structure.
The flow of the method does not specifically limit the architecture of the pre-trained artificial intelligence model, as long as the model can output the probability that each edge belongs to the weight and the minimum path in the graph structure according to the input graph structure.
In an alternative embodiment, the pre-trained artificial intelligence model may be a graph-convolution neural network model.
The atlas neural network model may include an input layer, an atlas layer, a fully connected layer, and a softmax layer.
Specifically, a graph structure may be input into a graph convolution neural network model, a characterization vector (embedding) of each edge in the graph structure is obtained through a graph convolution layer, then, a probability of belonging to a weight and a minimum path in the graph structure may be predicted for each edge through a full connection layer, and model training may be performed with a difference between a probability of belonging to the weight and the minimum path in the graph structure (i.e., an optimal solution of a traveler problem) of each edge output by a softmax layer and a case of an actual weight and a minimum path in the graph structure as a loss.
Because m is usually small (the number of nodes of a sub-graph is usually small), the super parameters of the graph convolution neural network model, such as the number of graph convolution layers, the number of full-connection layers, the number of neurons in a hidden layer, and the like, can be set to be small, so that the computing resources can be saved, and the computing efficiency can be improved.
3. And aiming at each edge in the task relationship graph, acquiring the weight and the minimum path probability of the edge in each sub-graph, belonging to the sub-graph, and integrating all the acquired probabilities to determine the weight and the minimum path probability of the edge belonging to the task relationship graph.
The embodiment does not limit the method for determining the probability of the edge in the task relationship graph according to the prediction result of the subgraph.
In an alternative embodiment, the weights and minimum paths in the subgraph of the task relationship graph may be part of the weights and minimum paths in the task relationship graph. Therefore, the probability that each edge belongs to the weight and the minimum path in the task relationship graph can be determined by referring to the probability that each edge belongs to each subgraph and the weight and the minimum path of each edge in each subgraph.
Therefore, optionally, by an averaging method, for each edge in the task relationship graph, a sum of the weight of the edge in each sub-graph to which the edge belongs and the probability of the minimum path is calculated, and then an average of the probabilities is calculated, specifically, the calculated sum of the probabilities may be divided by the number of sub-graphs to which the edge belongs. The calculated average may then be determined as the probability that the edge belongs to the weight and the minimum path in the task relationship graph.
Alternatively, the calculation may be performed by other methods, for example, removing the maximum and minimum probability values and calculating the average value, or determining the weight according to the importance of the subgraph, calculating the weighted sum, or calculating the median, etc.
Alternatively, the determination may be made by calculating a weighted sum. Integrating all the obtained probabilities to determine the weight of the edge belonging to the task relationship graph and the probability of the minimum path, which may include: calculating the weighted sum of all the obtained probabilities based on the probability weights respectively corresponding to all the obtained probabilities; and determining the weighted sum obtained by calculation as the probability that the edge belongs to the weight and the minimum path of the task relational graph.
The probability weight in the subgraph is not limited in the embodiment, and optionally, if the number of edges in the subgraph is larger, the weight and the minimum path in the subgraph are more likely to belong to the weight and the minimum path in the task relationship graph. Especially for subgraphs of the same number of nodes.
Therefore, optionally, for the weight and the probability of the minimum path that the edge belongs to any sub-graph in any sub-graph, the corresponding probability weight may be positively correlated with the number of edges in the sub-graph.
In an alternative embodiment, the output of the artificial intelligence model may be in the form of a matrix.
For example, the task relationship graph includes n nodes, the sub-graphs obtained by partitioning the task relationship graph include m nodes, and the probability set output by the artificial intelligence model for each sub-graph may be a matrix X of m × m. Xij may represent the probability that the edge between node i and node j belongs to the weight and minimum path in the corresponding subgraph.
And integrating the probability sets of all the sub-graphs to obtain a probability set corresponding to the task relation graph, wherein the probability set can be a matrix Y of n x n. Yij may represent the probability that the edge between node i and node j belongs to the weight and the minimum path in the task relationship graph.
S103: and according to the determined probability of each edge, determining a path reaching the standard from all possible paths which do not repeatedly pass through each node in the task relation graph.
The probability sum of the qualifying paths is greater than at least one other path among all possible paths that do not repeatedly pass through the nodes of the task relationship graph.
In an alternative embodiment, through the artificial intelligence model in S102, the weight that each edge in the task relationship graph belongs to the task relationship graph and the probability of the minimum path may be determined, and for the determined probability, the path that reaches the standard may be determined by using a simple search strategy.
Optionally, for the determined probability, a highest path of all possible paths that do not repeatedly pass through each node of the task relationship graph may be solved. Obviously, the higher the probability sum, the more likely it is the actual weights and minimum paths in the task relationship graph.
Correspondingly, the qualifying path may be the highest one of the probability sums among all possible paths that do not repeatedly pass through the nodes of the task relationship graph. Thus, the probability sum of the qualifying paths is optionally maximized among all possible paths that do not repeatedly pass through the nodes of the task relationship graph.
Alternatively, the solution may be performed in a traversal manner.
However, in general, when the task relationship graph is complex, it is often difficult to traverse all possible paths that do not repeatedly pass through each node of the task relationship graph.
For example, it is first necessary to traverse each node in the task relationship graph as a starting node to determine whether there are all possible paths from the starting node that do not repeatedly pass through the nodes of the task relationship graph.
Obviously, when the number of nodes in the task relationship graph is large, traversal is often difficult.
Alternatively, a path with a relatively large probability and a relatively large probability may be determined as the path reaching the standard from all possible paths that do not repeatedly pass through each node of the task relationship graph. Specifically, the local optimal solution may be used.
Thus, the probability sum of the qualifying path is greater than at least one other path among all possible paths that do not repeatedly pass through the nodes of the task relationship graph.
The embodiment does not limit the specific way of determining the standard-reaching path, and optionally, a greedy search strategy or a bundle search strategy may be adopted as an initial node for any node in the task relationship graph to determine a path with a relatively high probability and a relatively high probability as the standard-reaching path.
In an alternative embodiment, after determining the probability of each edge in the task relationship graph, it is further required to determine the path reaching the standard by using one node as a starting node.
The present embodiment does not limit the method of specifically selecting the start node.
Alternatively, a node may be randomly selected from the task relationship graph, or a node may be specified from the task relationship graph.
Alternatively, the selection may be made from nodes in the task relationship graph having a smaller or smallest number of connected edges. For example, a node that connects only one edge in the task relationship graph may be determined as a start node.
It should be noted that, for the selected start node, at least two paths that do not repeatedly pass through all nodes exist in the task relationship graph, so that the selection is facilitated.
The embodiment does not limit the specific method for determining the path reaching the standard in the task relationship diagram.
Alternatively, starting from the start node, in all edges connected by the start node, one edge with the highest corresponding probability is selected and determined as a part of the standard reaching path.
The corresponding probability is the probability that the edge belongs to the weight and the minimum path of the task relationship graph.
Then, another node connected by the selected edge can be further determined, and further for another node, all edges (except the selected edge) connected by another node are continuously selected, and the edge with the highest corresponding probability is selected to be determined as a part of the standard reaching path.
The above process is repeated until each node in the task relationship graph is passed by the qualifying path.
The processing sequence among the acquired tasks can be directly determined according to the standard reaching path.
Of course, in the process of determining the standard reaching path, as a part of the standard reaching path is determined, the processing sequence between the corresponding tasks can be directly determined, so that the processing sequence between the tasks can be determined in real time.
In the above method of determining the optimal path, a greedy search strategy is used, and one edge having the highest correspondence probability is selected from all edges (except for the selected edge) connected to one node.
Alternatively, other search strategies, such as beam search, etc., may be used.
In an alternative embodiment, determining the qualifying path from all possible paths that do not repeatedly pass through nodes in the task relationship graph based on the determined probability of each edge may include the following steps.
And selecting any node from the task relation graph to be added to the standard reaching path, and determining the node as the current node.
And circularly executing the following steps until each node in the task relation graph is added to the reach-to-standard path.
And determining the neighbor nodes which are not added into the standard reaching path in the neighbor nodes of the current node. Here to avoid duplicate nodes in the path.
In case the determined neighbor node comprises one, the edge between the current node and the neighbor node, are added to the reach-up path.
Adding edges between the current node and a preset neighbor node and the preset neighbor node to a standard reaching path under the condition that the determined neighbor nodes comprise at least two nodes; in the determined plurality of neighbor nodes, the probability of an edge between a neighbor node and the current node is preset to be at least greater than the probability of an edge between another neighbor node and the current node.
In this embodiment, each time when selecting a subsequent node, a next node corresponding to an edge with a relatively high probability is usually selected, so that a path with a relatively high probability and a relatively high probability can be determined.
Optionally, in the determined multiple neighbor nodes, the probability of an edge between a preset neighbor node and the current node may be greater than the probability of an edge between each other neighbor node and the current node. The embodiment may specifically be a greedy search strategy.
Optionally, in a case that the task relationship graph is a directed graph, the determined neighbor nodes may include a neighbor node pointed to by a directed edge from the current node.
Of course, alternatively, different nodes in the task relationship graph may be used as the starting points, and a plurality of preselected paths are obtained through the search strategy, so that the path with the highest probability and the lowest weight or the path with the lowest weight and the highest probability is selected.
Therefore, in an alternative embodiment, determining the qualifying path from all possible paths that do not repeatedly pass through nodes in the task relationship graph based on the determined probability of each edge may include the following steps.
And selecting a plurality of nodes from the task relation graph, and determining alternative paths according to a preset method.
And selecting the candidate path with the highest probability and the highest weight or the smallest weight from the plurality of determined candidate paths to be the standard reaching path. The probability sum or the weight sum of the qualifying paths in this embodiment may be at least greater than one other path.
Wherein at least two nodes may be selected from the task relationship graph.
Optionally, the preset method may specifically include: the currently selected node is added to the alternate path and determined to be the current node.
And circularly executing the following steps until each node in the task relation graph is added to the alternative path.
And determining neighbor nodes which are not added into the alternative path in the neighbor nodes of the current node.
In case the determined neighboring node comprises one, the edge between the current node and the neighboring node, and the neighboring node are added to the alternative path.
Adding edges between the current node and a preset neighbor node and the preset neighbor node to the alternative path under the condition that the determined neighbor nodes comprise at least two; in the determined plurality of neighbor nodes, the probability of an edge between a neighbor node and the current node is preset to be at least greater than the probability of an edge between another neighbor node and the current node.
In an alternative embodiment, determining the qualifying path from all possible paths that do not repeatedly pass through nodes in the task relationship graph based on the determined probability of each edge may include the following steps.
And selecting a node from the task relation graph to determine the node as the current node.
And circularly executing the following steps until each node in the task relation graph is added to the reach path.
And in the task relation graph, selecting a node which has an edge with the current node and is not added to the standard reaching path. And may specifically be all nodes.
Edges in the task relationship graph connecting the current node and each of the selected nodes are determined.
And selecting one edge meeting the preset probability requirement from all the determined edges.
And adding other nodes connected by the selected edge and the selected edge to the standard reaching path, and determining the other nodes as the current nodes.
Currently, in the case that the selected node has only one node, the edge between the current node and the node may be directly added to the reach path, and the node may be determined as the current node.
Optionally, in the greedy search strategy method, the edge meeting the preset probability requirement may be an edge with the highest probability of belonging to the weight and the minimum path in the task relationship graph among all the determined edges.
Optionally, in the method of the beam search strategy, selecting one edge meeting the preset probability requirement from all the determined edges may include the following steps.
Determining other nodes connected with the current node through any preset loop-free path in the task relation graph; the preset loop-free path may include a preset number of edges and any one of the determined edges; and the preset loop-free path may be a path that does not contain a loop.
Wherein the probability sum of each preset loop-free path can be calculated; and determining the edge connected with the current node as the edge meeting the requirement of the preset probability in the calculated probability and the maximum preset loop-free path.
It can be seen that when the number is preset to be 2 or more, the subsequent probability situation can be better viewed.
Alternatively, loop-free paths may be used to help determine the weights and minimum paths, since the weights and minimum paths do not repeatedly traverse the same node, and thus no loops exist.
Correspondingly, since all the determined edges are selected, the preset loop-free path needs to include any one of the determined edges, so that subsequent selection is facilitated.
The preset loop-free path may include a preset number of edges, and specifically may include one or more edges.
In the case that the preset loop-free path includes one edge, this embodiment may be regarded as a greedy search strategy.
In the case that the preset loop-free path includes a plurality of edges, calculating the sum of the probabilities may help determine the preset loop-free path having the largest sum of the probabilities in the case of a plurality of consecutive edges, so that the probability of determining the maximum path and the probability of determining the maximum path may be improved.
In this embodiment, based on the weight of each edge in the task relationship graph and the probability of the minimum path, the path up to standard in the task relationship graph can be determined through a simple search strategy, so that the efficiency of determining the path up to standard is improved, and the efficiency of determining the task processing sequence is also improved.
For ease of understanding, a specific example is provided below.
The form of the weight and the minimum path probability that each edge in the task relational graph belongs to the task relational graph may be a matrix, specifically, the matrix Z of n × n may be obtained for a task relational graph including n nodes, and Zij may represent the probability that the edge between the node i and the node j belongs to the weight and the minimum path in the task relational graph.
After matrix Z is obtained, a qualifying path may need to be obtained therefrom by a simple search strategy.
Specifically, a node i can be randomly selected as a starting point, the node i is added to the standard-reaching path, and an edge with the highest corresponding probability is selected from edges between the node i and neighbor nodes of the node i in a greedy manner.
Specifically, all of Zi1, Zi2, Zi3,. and Zin are determined from the matrix Z, and the highest probability, for example Zix, is selected, and then the edge between node x and node i, and node x may be added to the path to reach the standard.
Thereafter, for node x, the edges between node x and the neighbor nodes of node x may continue to be selected "greedily" with the highest corresponding probability.
This process may then be repeated until it is determined that n nodes are all in the qualifying path.
S104: and determining the processing sequence of each task according to the connection relation of each task in the standard-reaching path.
In the standard reaching path, the connection relation among all the tasks is clear, so that the processing sequence among all the tasks can be determined according to the connection relation.
Optionally, when the task relationship graph is a directed graph, any directed edge in the path reaching the standard, for example, a directed edge in which the first task points to the second task, may represent a processing order in which the first task and the second task are successively processed.
Optionally, in a case that the task relationship graph is an undirected graph, any undirected edge in the reach path, for example, an undirected edge in which a first task points to a second task, may characterize a processing order of continuously processing the first task and the second task. The order of precedence is not limited.
Since the standard reaching path does not repeatedly pass through each node in the task relationship graph, the processing order can be determined based on the standard reaching path for all tasks needing to determine the processing order in S101.
Because the probability sum of the paths reaching the standard is at least greater than that of other paths in all possible paths which do not repeatedly pass through each node of the task relation graph, the probability sum is relatively high, the weight sum of the paths reaching the standard is relatively small, the total time saved by the determined task processing sequence is relatively long, the total time saved by processing multiple tasks can be increased, the overall processing efficiency is improved, the task processing sequence with relatively high processing efficiency is determined, and the task processing is convenient to process.
The standard-reaching path is determined from the task relation graph based on the data analysis of the time saving among the tasks and aiming at increasing the total time saving, so that the total time saving of the multi-task processing can be improved to a certain extent and the overall processing efficiency of the multi-task processing can be improved due to the task processing sequence represented by the standard-reaching path.
Corresponding to the method embodiment, the embodiment of the present specification further provides an apparatus embodiment.
Fig. 4 is a schematic structural diagram of a multitask processing order determining apparatus according to an embodiment of the present specification. The apparatus may include the following elements.
An obtaining unit 401 is configured to obtain a task relationship diagram.
Wherein, different tasks needing to determine the processing sequence are taken as different nodes in the task relation graph; the weight of each edge in the task relation graph is inversely related to the saved time corresponding to the two tasks connected with the edge; the calculation mode of the saved time corresponding to the first task and the second task comprises the following steps: the duration of processing only the first task in one processing, plus the duration of processing only the second task in another processing, and minus the duration of processing both tasks consecutively in one processing.
And a probability unit 402, configured to determine, based on the artificial intelligence model, a weight and a probability of a minimum path that each edge in the task relationship graph belongs to the task relationship graph.
The method comprises the steps that an artificial intelligence model is trained in advance according to a graph sample and a label corresponding to each edge in the graph sample, wherein the label corresponding to each edge is used for representing the weight and the minimum path of the edge which belongs to or does not belong to the graph sample; the weight and the minimum path of the graph are the paths with the minimum weight and the minimum path in all possible paths which do not repeatedly pass through each node of the graph;
and a path unit 403, configured to determine, according to the determined probability of each edge, a path up to standard from all possible paths that never repeatedly pass through each node in the task relationship graph.
The probability sum of the qualifying paths is greater than at least one other path among all possible paths that do not repeatedly pass through the nodes of the task relationship graph.
And a sequence unit 404, configured to determine a processing sequence of each task according to a connection relationship of each task in the standard-reaching path.
For an explanation of the above-described embodiments of the apparatus, reference is made to the explanation of the process flow.
Embodiments of the present specification further provide a computer device, which at least includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements a multitask processing order determination method when executing the program, that is, any one of the method embodiments described above.
Fig. 5 is a schematic diagram illustrating a more specific hardware structure of a computer device according to an embodiment of the present disclosure, where the device may include: a processor 1010, a memory 1020, an input/output interface 1030, a communication interface 1040, and a bus 1050. Wherein the processor 1010, memory 1020, input/output interface 1030, and communication interface 1040 are communicatively coupled to each other within the device via bus 1050.
The processor 1010 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an Application Specific Integrated Circuit (ASIC), or one or more Integrated circuits, and is configured to execute related programs to implement the technical solutions provided in the embodiments of the present disclosure.
The Memory 1020 may be implemented in the form of a ROM (Read Only Memory), a RAM (Random Access Memory), a static storage device, a dynamic storage device, or the like. The memory 1020 may store an operating system and other application programs, and when the technical solution provided by the embodiments of the present specification is implemented by software or firmware, the relevant program codes are stored in the memory 1020 and called to be executed by the processor 1010.
The input/output interface 1030 is used for connecting an input/output module to input and output information. The i/o module may be configured as a component in a device (not shown) or may be external to the device to provide a corresponding function. The input devices may include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output devices may include a display, a speaker, a vibrator, an indicator light, etc.
The communication interface 1040 is used for connecting a communication module (not shown in the drawings) to implement communication interaction between the present apparatus and other apparatuses. The communication module can realize communication in a wired mode (such as USB, network cable and the like) and also can realize communication in a wireless mode (such as mobile network, WIFI, Bluetooth and the like).
It should be noted that although the above-mentioned device only shows the processor 1010, the memory 1020, the input/output interface 1030, the communication interface 1040 and the bus 1050, in a specific implementation, the device may also include other components necessary for normal operation. In addition, those skilled in the art will appreciate that the above-described apparatus may also include only those components necessary to implement the embodiments of the present description, and not necessarily all of the components shown in the figures.
Embodiments of the present description further provide a computer-readable storage medium, on which a computer program is stored, which when executed by a processor implements a multitask order determining method, namely any of the above-described method embodiments.
Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, 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 technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer readable medium does not include a transitory computer readable medium such as a modulated data signal and a carrier wave.
From the above description of the embodiments, it is clear to those skilled in the art that the embodiments of the present disclosure can be implemented by software plus necessary general hardware platform. Based on such understanding, the technical solutions of the embodiments of the present specification may be embodied in the form of a software product, which may be stored in a storage medium, such as a ROM/RAM, a magnetic disk, an optical disk, or the like, and includes instructions for enabling a computer device (which may be a personal computer, a server, or a network device) to execute a multitask processing order determination method described in various embodiments or some parts of the embodiments of the present specification, that is, any method embodiment described above.
The systems, apparatuses, modules or units described in the above embodiments may be specifically implemented by a computer chip or an entity, or implemented by a product with certain functions. A typical implementation device is a computer, which may take the form of a personal computer, laptop computer, cellular telephone, camera phone, smart phone, personal digital assistant, media player, navigation device, email messaging device, game console, tablet computer, wearable device, or a combination of any of these devices.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the apparatus embodiment, since it is substantially similar to the method embodiment, it is relatively simple to describe, and reference may be made to some descriptions of the method embodiment for relevant points. The above-described apparatus embodiments are merely illustrative, and the modules described as separate components may or may not be physically separate, and the functions of the modules may be implemented in one or more software and/or hardware when implementing the embodiments of the present disclosure. And part or all of the modules can be selected according to actual needs to achieve the purpose of the scheme of the embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
The foregoing is only a detailed description of the embodiments of the present disclosure, and it should be noted that, for those skilled in the art, many modifications and decorations can be made without departing from the principle of the embodiments of the present disclosure, and these modifications and decorations should also be regarded as protection for the embodiments of the present disclosure.
Claims (10)
1. A multitasking order determination method comprising:
acquiring a task relation graph; wherein, different tasks needing to determine the processing sequence are taken as different nodes in the task relation graph; the weight of each edge in the task relation graph is inversely related to the saved time corresponding to the two tasks connected with the edge; the calculation mode of the saved time corresponding to the first task and the second task comprises the following steps: a time length for processing only the first task in one processing, a time length for processing only the second task in another processing, and a time length for continuously processing the two tasks in one processing are subtracted;
determining the weight of each edge in the task relation graph and the probability of the minimum path based on an artificial intelligence model; training an artificial intelligence model in advance according to the graph sample and a label corresponding to each edge in the graph sample, wherein the label corresponding to each edge is used for representing the weight and the minimum path of the edge which belongs to or does not belong to the graph sample; the weight and the minimum path of the graph are the paths with the minimum weight and the minimum path in all possible paths which do not repeatedly pass through each node of the graph;
determining a path reaching the standard from all possible paths which never repeatedly pass through each node in the task relation graph according to the determined probability of each edge; the probability sum of the qualifying paths is at least greater than one other path among all possible paths that do not repeatedly pass through nodes of the task relationship graph;
and determining the processing sequence of each task according to the connection relation of each task in the standard reaching path.
2. The method of claim 1, wherein determining the probability that each edge in the task relationship graph belongs to the weight and the minimum path of the task relationship graph based on the artificial intelligence model comprises:
splitting a plurality of subgraphs aiming at the task relation graph; each node in the task relation graph at least belongs to one subgraph;
inputting each sub-graph into an artificial intelligence model, and outputting the weight and the minimum path probability of each edge in the sub-graph, wherein each edge belongs to the sub-graph;
and aiming at each edge in the task relationship graph, acquiring the weight and the minimum path probability of the edge in each sub-graph, belonging to the sub-graph, and integrating all the acquired probabilities to determine the weight and the minimum path probability of the edge belonging to the task relationship graph.
3. The method of claim 2, the splitting out a plurality of subgraphs for a task relationship graph, comprising:
aiming at the task relationship graph, the following steps are executed in a circulating mode until each node in the task relationship graph at least belongs to one sub graph:
selecting a node which does not belong to any subgraph in the task relationship graph, and determining the node as a central node;
determining other nodes meeting the preset screening condition as alternative nodes in other nodes except the central node;
for each alternative node, determining the path weight sum from the central node to the alternative node;
according to the determined weight sum, sequencing the alternative nodes in a sequence from small to large; determining the alternative nodes with the sequence numbers smaller than the preset sequencing sequence number in the sequencing result as the sub-graph nodes;
and forming a subgraph by the central node, the determined subgraph nodes and the path from the central node to each subgraph node.
4. The method of claim 2, wherein all the pattern book nodes for training the artificial intelligence model are the same in number;
splitting a plurality of sub-graphs aiming at the task relationship graph, wherein the splitting comprises the following steps: splitting a plurality of sub-graphs containing a preset number of nodes aiming at the task relation graph; the preset node number is the node number of the graph sample.
5. The method of claim 1, wherein determining an up-to-standard path from all possible paths that never repeatedly pass through nodes in the task relationship graph according to the determined probability of each edge comprises:
selecting any node from the task relation graph to be added to the standard reaching path, and determining the node as a current node;
circularly executing the following steps until each node in the task relation graph is added to the standard reaching path:
determining neighbor nodes which are not added into the standard reaching path in neighbor nodes of the current node;
in the case that the determined neighbor node includes one, adding an edge between the current node and the neighbor node, and the neighbor node to the reach path;
adding edges between the current node and preset neighbor nodes and the preset neighbor nodes to the standard reaching path under the condition that the determined neighbor nodes comprise at least two; in the determined plurality of neighbor nodes, the probability of an edge between the preset neighbor node and the current node is at least greater than the probability of an edge between another neighbor node and the current node.
6. The method of claim 5, wherein the determined plurality of neighboring nodes have a probability of an edge between the predetermined neighboring node and the current node that is greater than a probability of an edge between each of the other neighboring nodes and the current node.
7. The method of claim 1, the probability sum of the qualifying paths being the largest of all possible paths that do not repeatedly pass through nodes of a task relationship graph.
8. A multitasking order determining apparatus comprising:
the acquiring unit is used for acquiring a task relation graph; wherein, different tasks needing to determine the processing sequence are taken as different nodes in the task relation graph; the weight of each edge in the task relation graph is inversely related to the saved time corresponding to the two tasks connected with the edge; the calculation mode of the saved time corresponding to the first task and the second task comprises the following steps: a time length for processing only the first task in one processing, a time length for processing only the second task in another processing, and a time length for continuously processing the two tasks in one processing are subtracted;
the probability unit is used for determining the weight of each edge in the task relation graph and the probability of the minimum path based on the artificial intelligence model; training an artificial intelligence model in advance according to the graph sample and a label corresponding to each edge in the graph sample, wherein the label corresponding to each edge is used for representing the weight and the minimum path of the edge which belongs to or does not belong to the graph sample; the weight and the minimum path of the graph are the paths with the minimum weight and the minimum path in all possible paths which do not repeatedly pass through each node of the graph;
the path unit is used for determining a path reaching the standard from all possible paths which never repeatedly pass through each node in the task relation graph according to the determined probability of each edge; the probability sum of the qualifying paths is greater than at least one other path among all possible paths that do not repeatedly pass through nodes of the task relationship graph;
and the sequence unit is used for determining the processing sequence of each task according to the connection relation of each task in the standard reaching path.
9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method of any one of claims 1 to 7 when executing the program.
10. A computer-readable storage medium, the storage medium storing a computer program for implementing the method of any one of claims 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210489685.5A CN114841664A (en) | 2022-05-06 | 2022-05-06 | Method and device for determining multitasking sequence |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210489685.5A CN114841664A (en) | 2022-05-06 | 2022-05-06 | Method and device for determining multitasking sequence |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114841664A true CN114841664A (en) | 2022-08-02 |
Family
ID=82568494
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210489685.5A Pending CN114841664A (en) | 2022-05-06 | 2022-05-06 | Method and device for determining multitasking sequence |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114841664A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116795519A (en) * | 2023-08-25 | 2023-09-22 | 江苏盖睿健康科技有限公司 | Internet-based remote intelligent debugging method and system |
-
2022
- 2022-05-06 CN CN202210489685.5A patent/CN114841664A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116795519A (en) * | 2023-08-25 | 2023-09-22 | 江苏盖睿健康科技有限公司 | Internet-based remote intelligent debugging method and system |
CN116795519B (en) * | 2023-08-25 | 2023-12-05 | 江苏盖睿健康科技有限公司 | Internet-based remote intelligent debugging method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109242135B (en) | Model operation method, device and business server | |
EP3574453A1 (en) | Optimizing neural network architectures | |
CN111435482A (en) | Outbound model construction method, outbound method, device and storage medium | |
KR102330423B1 (en) | Online default forecasting system using image recognition deep learning algorithm | |
CN112700006B (en) | Network architecture searching method, device, electronic equipment and medium | |
CN111815169A (en) | Business approval parameter configuration method and device | |
CN111652661B (en) | Mobile phone client user loss early warning processing method | |
CN113205403A (en) | Method and device for calculating enterprise credit level, storage medium and terminal | |
CN112884569A (en) | Credit assessment model training method, device and equipment | |
CN110309947A (en) | Complete vehicle logistics order forecast method and device, logistics system and computer-readable medium | |
CN114428748B (en) | Simulation test method and system for real service scene | |
CN113268403A (en) | Time series analysis and prediction method, device, equipment and storage medium | |
CN110263136B (en) | Method and device for pushing object to user based on reinforcement learning model | |
CN114841664A (en) | Method and device for determining multitasking sequence | |
CN115130887A (en) | Reservoir dam environmental impact evaluation method and device, electronic equipment and storage medium | |
CN113268485B (en) | Data table association analysis method, device, equipment and storage medium | |
CN114238135A (en) | Test case determination method and device and electronic equipment | |
CN114154622A (en) | Algorithm model for traffic operation system flow data acquisition missing completion | |
Bevacqua et al. | A data-adaptive trace abstraction approach to the prediction of business process performances | |
CN111783487B (en) | Fault early warning method and device for card reader equipment | |
CN112015426A (en) | Code management method, device and equipment | |
CN110309948A (en) | Complete vehicle logistics order forecast method and device, logistics system and computer-readable medium | |
CN111950753A (en) | Scenic spot passenger flow prediction method and device | |
CN115204501A (en) | Enterprise evaluation method and device, computer equipment and storage medium | |
CN111325401B (en) | Method and device for training path planning model and computer system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |