Task execution method, device and equipment

Abstract

The embodiment of the invention discloses a task execution method, device and equipment. According to the scheme, the method comprises the steps that after sequence operation information, related to a target resource, of a first user in a first time period is obtained, the sequence operation information, a time interval where the obtaining moment of the sequence operation information is located and m candidate tasks are input into a neural network model, and evaluation results, output by the neural network model, of the m candidate tasks are obtained; wherein the evaluation result is used for reflecting the probability that the first user obtains the target resource after the candidate task is executed on the first user; according to the evaluation result, a selected task used for being executed on the first user is determined from the m candidate tasks; wherein the probability of acquiring the target resource by the first user corresponding to the selected task is higher than that of other candidate tasks; finally, the selected task is performed on the first user within a time interval.

Description

Technical Field

[0001] This application relates to the technical field of task processing, and particularly to a task execution method, apparatus, and device. Background Art

[0002] With the development of the Internet, marketing means such as issuing coupons, awarding prizes, and reducing prices have become increasingly important in marketing tasks.

[0003] In related technologies, marketing information is usually recommended to users based on the users' historical data. However, the historical data of users often has low timeliness, resulting in the marketing information recommended to users usually not meeting the current business needs of users, insufficient accuracy of information recommendation, low user conversion rate, and easy waste of marketing resources.

[0004] In view of this, a task execution solution with high timeliness and high accuracy is required. Summary of the Invention

[0005] In view of this, embodiments of this application provide a task execution method, apparatus, and device to provide a task execution solution with high timeliness and high accuracy.

[0006] To solve the above technical problems, an embodiment of this specification provides a task execution method, including:

[0007] Obtain sequence operation information related to a target resource of a first user within a first time period; the time interval between the end time of the first time period and the acquisition time of the sequence operation information is less than or equal to a first duration threshold; the first duration threshold is at the hour level or the minute level; the sequence operation information includes a number of operation information sorted by occurrence time;

[0008] Input the sequence operation information, the time interval where the acquisition time of the sequence operation information is located, and m candidate tasks into a neural network model, and obtain an evaluation result of the m candidate tasks output by the neural network model; the time interval is at the hour level or the minute level; m is an integer greater than or equal to 2; the evaluation result is used to reflect the probability that the first user obtains the target resource after the candidate task is executed for the first user;

[0009] Determine a selected task to be executed for the first user from the m candidate tasks according to the evaluation result; the probability that the first user corresponding to the selected task obtains the target resource is higher than that of other candidate tasks;

[0010] Execute the selected task for the first user within the time interval.

[0011] An embodiment of this specification also provides a task execution apparatus, including:

[0012] An information acquisition module, configured to acquire sequence operation information of a first user related to a target resource within a first time period; a time interval between an end moment of the first time period and an acquisition moment of the sequence operation information is less than or equal to a first duration threshold; the first duration threshold is at the hour level or the minute level; the sequence operation information includes a plurality of operation information sorted according to occurrence time;

[0013] A model inference module, configured to input the sequence operation information, a time interval where the acquisition moment of the sequence operation information is located, and m candidate tasks into a neural network model, and obtain an evaluation result of the m candidate tasks output by the neural network model; the time interval is at the hour level or the minute level; m is an integer greater than or equal to 2; the evaluation result is used to reflect a probability that the first user acquires the target resource after the candidate task is executed for the first user;

[0014] A task selection module, configured to determine a selected task to be executed for the first user from the m candidate tasks according to the evaluation result; a probability that the first user corresponding to the selected task acquires the target resource is higher than that of other candidate tasks;

[0015] A task execution module, configured to execute the selected task for the first user within the time interval.

[0016] An embodiment of this specification further provides a task execution device, including:

[0017] At least one processor; and,

[0018] A memory communicatively connected to the at least one processor; wherein,

[0019] The memory stores instructions executable by the at least one processor, and when the instructions are executed by the at least one processor, the at least one processor is enabled to:

[0020] Acquire sequence operation information of a first user related to a target resource within a first time period; a time interval between an end moment of the first time period and an acquisition moment of the sequence operation information is less than or equal to a first duration threshold; the first duration threshold is at the hour level or the minute level; the sequence operation information includes a plurality of operation information sorted according to occurrence time;

[0021] Input the sequence operation information, the time interval in which the acquisition time of the sequence operation information is located, and m candidate tasks into a neural network model to obtain the evaluation results of the m candidate tasks output by the neural network model; the time interval is at the hour level or the minute level; m is an integer greater than or equal to 2; the evaluation results are used to reflect the probability that the first user obtains the target resource after performing the candidate task on the first user;

[0022] According to the evaluation results, determine the selected task to be performed on the first user from the m candidate tasks; the probability that the first user corresponding to the selected task obtains the target resource is higher than that of other candidate tasks;

[0023] Perform the selected task on the first user within the time interval.

[0024] At least one embodiment provided in this specification can achieve the following beneficial effects:

[0025] In the embodiments of this specification, after obtaining the sequence operation information related to the target resource of the first user in the first time period, input the sequence operation information, the time interval in which the acquisition time of the sequence operation information is located, and m candidate tasks into a neural network model, and the evaluation results of the m candidate tasks output by the neural network model can be obtained, and select the candidate task with the highest probability that the first user obtains the target resource as the selected task, and perform the selected task on the first user within the time interval. Among them, the acquisition time of the sequence operation information can correspond to the current time. Therefore, when the neural network model performs task inference, it needs to consider the time interval in which the current time is located; and when the neural network model performs task inference, the acquisition time of the user's sequence operation information it depends on is not far from the current time; furthermore, after the neural network model performs task inference, it needs to perform the selected task with the highest probability selected based on the inference result of the model within the time interval in which the current time is located. To sum up, on the one hand, by referring to the user's sequence operation information that is not far from the current time in a timely manner, the timeliness of the inference result output by the model itself is relatively high. In other words, it ensures that the inference result output by the model has high timeliness; on the other hand, by referring to the time interval in which the current time is located when obtaining the inference result, and timely performing the selected task selected based on the inference result of the model within the time interval in which the current time is located, the high timeliness of task execution is guaranteed. In addition, in the embodiments of this specification, by real-time sensing the user's sequence operation information and making real-time decisions, the characteristics of the user can be grasped in a timely manner, and the user's needs can be mined in real time through deep learning algorithms, so that the selected task determined based on the neural network model can meet the current personalized needs of the user, which is beneficial to improving the accuracy of task execution. Brief Description of the Drawings

[0026] To more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments recorded in this application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0027] Figure 1 Schematic diagram of an application scenario of a task execution method provided by an embodiment of this specification;

[0028] Figure 2 Flowchart of a task execution method provided by an embodiment of this specification;

[0029] Figure 3 Flowchart of a population screening and task execution method provided by an embodiment of this specification;

[0030] Figure 4 Flowchart of a neural network model training method provided by an embodiment of this specification;

[0031] Figure 5 Flowchart of a sample scheme construction method provided by an embodiment of this specification;

[0032] Figure 6 Corresponding to an embodiment of this specification Figure 2 Structural schematic diagram of a task execution device;

[0033] Figure 7 Corresponding to an embodiment of this specification Figure 2 Structural schematic diagram of a task execution device. Detailed implementation manners

[0034] Many specific details are set forth in the following description to facilitate a thorough understanding of this application. However, this application can be implemented in many other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the spirit of this application. Therefore, this application is not limited by the specific implementations disclosed below.

[0035] The terms used in one or more embodiments of this application are only for the purpose of describing specific embodiments and are not intended to limit one or more embodiments of this application. The singular forms "a", "the", and "said" used in one or more embodiments of this application and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used in one or more embodiments of this application refers to and includes any or all possible combinations of one or more of the associated listed items.

[0036] It should be understood that although the terms first, second, etc. may be used in one or more embodiments of the present application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from each other. For example, without departing from the scope of one or more embodiments of the present application, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "when" or "while" or "in response to determining".

[0037] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of the relevant data need to comply with the relevant laws, regulations, and standards of the relevant region, and corresponding operation entrances are provided for the user to select authorization or rejection.

[0038] In the related art, the general steps of the traditional marketing plan recommendation are as follows: when an operator decides on a marketing plan, first, based on expert experience or data such as the user's personal information and historical transaction information, an algorithm recommendation score is made, a part of users is selected, and this marketing plan is launched for this part of users. However, in some fields with high requirements for timeliness, such as the consumer credit field, different from traditional industries such as e-commerce, user needs often do not have high frequency and periodicity, and historical data often makes the recommendation outdated more easily, unable to capture the user's current business needs, resulting in low timeliness, and the recommendation result may not match the user's current needs, resulting in poor accuracy.

[0039] To solve the defects in the related art, the following embodiments are given in this solution.

[0040] Figure 1 It is a schematic diagram of an application scenario of a task execution method provided in an embodiment of this specification.

[0041] Such as Figure 1As shown in the figure, the terminal device 101 of the first user can send the sequence operation information related to the target resource by the first user within the first time period to the task execution platform server 102. After receiving the sequence operation information, the task execution platform server 102 can input the sequence operation information, the time interval where the acquisition time of the sequence operation information is located, and m candidate tasks into the neural network model obtained from the model training server 103, obtain the evaluation results of the m candidate tasks output by the neural network model, and according to the evaluation results, determine the task with the highest probability for the first user to obtain the target resource from the m candidate tasks as the selected task, and execute the selected task for the first user within the time interval.

[0042] Among them, the model training server 103 can be a server for training a neural network model. The model training server 103 can obtain the sample data stored in the database 104 from the database 104, and use the sample data to train and update the neural network model. In practical applications, the database 104 can obtain data from the user terminal in real time and make it into sample data, and send these sample data to the model training server 103 in real time, so that the model training server 103 can use the obtained sample data to update the neural network model in real time to improve the accuracy of the neural network model.

[0043] In practical applications, the task execution platform server 102 can interact with multiple terminal devices. After the task execution platform server 102 obtains the sequence operation information uploaded in real time by the user's terminal device, it can use the neural network model to process these sequence operation information in real time, and determine the task with the highest probability for the user to obtain the target resource within the time interval where the current moment is located, and execute the selected task in real time to ensure the timeliness and accuracy of task execution.

[0044] Among them, the terminal device 101 of the first user includes but is not limited to smart phones, tablet computers, palm computers, smart watches, etc. The task execution platform server 102 includes but is not limited to any device, equipment, platform, server cluster, etc. with computing and processing capabilities. The model training server 103 includes but is not limited to any device, equipment, platform, server cluster, etc. with computing and processing capabilities. The database 104 includes but is not limited to systems or devices with data storage functions such as relational databases, non-relational databases, graph databases, columnar databases, etc.

[0045] In practical applications, between the terminal device 101 of the first user and the task execution platform server 102, between the task execution platform server 102 and the model training server 103, and between the model training server 103 and the database 104, data transmission can be carried out through local area network connection, wide area network connection, Internet connection or other types of data network connections, or through other means, and specific limitations are not made in this regard.

[0046] In addition, although in at least some embodiments of this specification as Figure 1 shown, in the task execution method provided by the embodiments of this specification, the model training server 103 obtains sample data from the database 104, however, in other alternative embodiments, the model training server 103 can also directly obtain sample data from the user terminal or other devices. In addition, in other alternative embodiments, the task execution platform server 102 can also directly train the neural network model, thereby avoiding the situation of affecting timeliness due to the slow model data transmission speed.

[0047] Figure 1 In the method in, after the task execution platform obtains the sequence operation information related to the target resource of the first user within the first time period, the sequence operation information, the time interval where the acquisition moment of the sequence operation information is located, and m candidate tasks are input into the neural network model, and the evaluation results of these m candidate tasks output by the neural network model can be obtained, and the candidate task with the highest probability for the first user to obtain the target resource is selected as the selected task, and the selected task is executed for the first user within the time interval. The acquisition moment of the sequence operation information can correspond to the current moment. Thus, when the neural network model performs task inference, it needs to consider the time interval where the current moment is located; and when the neural network model performs task inference, the acquisition time of the user's sequence operation information on which it is based is not far from the current moment; furthermore, after the neural network model performs task inference, it needs to execute the selected task with the highest probability selected based on the inference result of the model within the time interval where the current moment is located. In summary, on the one hand, by timely referring to the user's sequence operation information not far from the current moment, the timeliness of the inference result output by the model itself is relatively high. In other words, it is ensured that the inference result output by the model has high timeliness; on the other hand, by referring to the time interval where the current moment is located when obtaining the inference result, and timely executing the selected task selected based on the inference result of the model within the time interval where the current moment is located, the high timeliness of task execution is guaranteed. In addition, by real-time perceiving the user's sequence operation information and making real-time decisions, the user's characteristics can be grasped in a timely manner, and the user's needs can be mined in real time through deep learning algorithms, so that the selected task determined based on the neural network model can meet the user's current personalized needs, which is beneficial to improving the accuracy of task execution.

[0048] Figure 2 This is a schematic flowchart of a task execution method provided by an embodiment of this specification. From a hardware perspective, the execution subject of this process can be a server. From a program perspective, the execution subject of this process can be an application program installed on the server. As Figure 2 shown, this process may include the following steps:

[0049] Step 202: Obtain sequence operation information of a first user related to a target resource within a first time period; the time interval between the end time of the first time period and the acquisition time of the sequence operation information is less than or equal to a first duration threshold; the first duration threshold is at the hour level or the minute level; the sequence operation information includes several pieces of operation information sorted by occurrence time.

[0050] In the embodiments of this specification, the target resource can be a resource or information that a user can obtain. When a user obtains a target resource, for example: the user accesses or purchases a certain credit product, the user downloads a certain software, the user purchases a certain commodity, the user accesses a certain page, etc. Taking the example of a user purchasing commodity A, commodity A can correspond to the target resource; taking the example of a user accessing web page B, the information in web page B can correspond to the target resource.

[0051] In practical applications, the sequence operation information can be sequence operation information having an association relationship with the target resource, and the sequence operation information can have a direct or indirect association relationship with the target resource. For example, the sequence operation information can be accessing the page or service where the target resource is located. Another example is that the sequence operation information can also be accessing the page or service where a resource of the same type as the target resource is located. Still another example is that the sequence operation information can also be accessing an associated page of the resource provider of the target resource, etc. Specifically, taking the marketing target that a user purchases a certain credit product as an example, the sequence operation information having an association relationship with the target resource may include, but is not limited to: browsing and clicking operations on this credit product, browsing and clicking operations on other credit products, browsing and clicking operations on any product of the company to which this credit product belongs, etc. By only obtaining and analyzing the sequence operation information related to the target resource, the amount of data that needs to be obtained and analyzed can be reduced, the waste of resources caused by obtaining and analyzing irrelevant data can be avoided, and it is beneficial to improve the execution efficiency of the solution.

[0052] In practical applications, the system of the user terminal device can collect the operation information of the user within a preset time period, and add a time stamp to the operation information for the moment when the operation occurs to mark the time point when each operation behavior occurs. The user terminal device or Figure 2After obtaining the operation information related to the target resource of the first user within the first time period, the execution subject of the method can sort these operation information according to the time stamps of each operation information in the order of occurrence time to obtain sequential operation information. The duration of the first time period is at the hour level or the minute level. In practical applications, it can be set and adjusted according to requirements. The first time period can be 10 minutes, 30 minutes, 1 hour, 3 hours, 12 hours, etc., and no specific limitation is made in this regard.

[0053] In the embodiment of this specification, the acquisition moment of the sequential operation information, that is, the current moment, which is also the execution moment of the task execution method of this application. Also, since the time interval between the end moment of the first time period and the acquisition moment (current moment) of the sequential operation information is less than or equal to the first duration threshold (the first duration threshold is at the hour level or the minute level). For example, the hour-level or minute-level first duration threshold can include 1 minute, several minutes, more than ten minutes, dozens of minutes, 1 hour, or several hours, etc. Therefore, the sequential operation information related to the target resource of the first user obtained is recent, is operation information close to the current moment, and has high timeliness. In practical applications, it can be set and adjusted according to requirements. The first duration threshold can be 1 minute, 10 minutes, 1 hour, etc., and no specific limitation is made in this regard. For example: the first duration threshold is set to 10 minutes, and the current moment is 9:30. It is possible to collect the sequential operation information related to the target resource of user A within the first 30-minute time period from 8:55 to 9:25, or it is also possible to collect the sequential operation information related to the target resource of user A within the first 2-hour time period from 7:20 to 9:20. Since the time difference between 9:25 and 9:30 is 5 minutes, and the time difference between 9:20 and 9:30 is 10 minutes, both meet the requirement of being less than or equal to the first duration threshold.

[0054] In addition, although the time information accurate to minutes is given as an example above, in practical applications, the time information corresponding to user operations can also be accurate to seconds or even finer granularity, which can be set according to actual needs.

[0055] Step 204: Input the sequential operation information, the time interval where the acquisition moment of the sequential operation information is located, and m candidate tasks into a neural network model to obtain the evaluation results of the m candidate tasks output by the neural network model; the time interval is at the hour level or the minute level; m is an integer greater than or equal to 2; the evaluation results are used to reflect the probability that the first user obtains the target resource after the candidate task is executed for the first user.

[0056] In the embodiments of this specification, a neural network model is a computational model that mimics the structure and functions of a biological neural network and is widely used in fields such as artificial intelligence, machine learning, and deep learning. The neural network model consists of a large number of artificial neurons (nodes), and these neurons are connected by weights and can learn and process complex data. The structure of the neural network model can generally be divided into an input layer, a hidden layer, and an output layer. Among them, the input layer is used to receive the original input data, the hidden layer is used to extract the features of the data, and the output layer is used to generate the final prediction result or classification label.

[0057] In the embodiments of this specification, the neural network model may include a neural network model capable of processing sequential data. Specifically, it may include but is not limited to: Convolutional Neural Networks (CNN) model, Recurrent Neural Network (RNN) model, Bidirectional Recurrent Neural Network (Bi-RNN), Long Short-Term Memory (LSTM) model, Gated Recurrent Unit (GRU) model, etc.

[0058] In actual application, the neural network model may be trained based on a sample scenario carrying sample labels. Among them, the sample scenario is used to represent that after a second user performs a sample sequence operation, a sample task is performed on the second user within a sample time interval. Among them, the sample task includes a first type of sample task or a second type of sample task; the first type of sample task is used to represent performing a resource recommendation operation for a sample resource to the user; the second type of sample task is used to represent not performing a resource recommendation operation for the sample resource to the user. Among them, the sample label includes a first sample label or a second sample label. The first sample label may represent that after performing the sample scenario on the second user, the second user obtains the sample resource; the second sample label may represent that after performing the sample scenario on the second user, the second user does not obtain the sample resource. The neural network model can learn the features of the data through the training process. After the neural network model is trained, it can, based on the user's sequential operation information, the time interval in which the acquisition moment of the sequential operation information is located, and a candidate task, predict the probability that the user obtains the target resource after performing the candidate task on the user within the time interval. The training process of the model generally may include steps such as forward propagation, loss calculation, backward propagation, and weight update. Since the subsequent embodiments in the embodiments of this specification will explain the training process of the model in detail, it will not be elaborated here.

[0059] In the embodiments of this specification, the time interval in which the acquisition time of the sequence operation information is located may refer to the time interval to which the current moment belongs, where the time interval is at the hour level or the minute level. For example, the time intervals at the hour level or the minute level may include 10 minutes, 30 minutes, 1 hour, 3 hours, etc. In practical applications, the time within a preset time period of a day may be divided into multiple time intervals according to actual needs. For example, to avoid disturbing users at night, the time period from 9:00 to 20:00 may be divided into 11 time intervals, each time interval being 1 hour, 9:00 - 10:00 being one time interval, 10:00 - 11:00 being one time interval, and so on. If the acquisition time of the sequence operation information is 9:45, then the time interval in which the acquisition time of the sequence operation information is located is the time interval of 9:00 - 10:00.

[0060] Step 206: Determine a selected task to be executed for the first user from the m candidate tasks according to the evaluation result; the probability that the first user corresponding to the selected task obtains the target resource is higher than that of other candidate tasks.

[0061] In practical applications, the m candidate tasks may be m solutions. For example: in a marketing scenario, the m candidate tasks may specifically be m marketing solutions. Specifically, among the m candidate tasks, each candidate task can form a set of input data of the model with the sequence operation information and the time interval in which the acquisition time of the sequence operation information is located, so as to obtain m sets of input data of the model. After inputting each set of input data into the neural network model, the model can output the evaluation result corresponding to the candidate task in this set of data. The evaluation result is used to reflect the probability that the first user obtains the target resource after the candidate task is executed for the first user.

[0062] In practical applications, among the m candidate tasks, each candidate task can correspond to an evaluation result, that is: each candidate task can correspond to a probability value. Thus, m probability values can be obtained. Specifically, the m probability values can be sorted according to the magnitudes of these m probability values, and the candidate task corresponding to the highest probability value ranked first can be determined as the selected task to be executed for the first user. Or, other methods can also be used to determine the highest probability value among the m probability values, and no specific limitation is made here.

[0063] Step 208: Execute the selected task for the first user within the time interval.

[0064] In the embodiments of this specification, the time interval is the time interval in which the acquisition time of the sequence operation information is located, and correspondingly, it is also the time interval in which the current time is located. Since the probability that the first user obtains the target resource after executing the selected task corresponding to the selected task predicted by the neural network model is calculated based on the time interval in which the acquisition time of the sequence operation information is located, therefore, executing the selected task on the first user within the time interval can ensure the timeliness and accuracy of task execution. For example: If the acquisition time of the sequence operation information is 9:45, and the time interval in which the acquisition time of the sequence operation information is located is the time interval from 9:00 to 10:00, then the selected task needs to be executed on the first user within the time interval from 9:00 to 10:00 to ensure the timeliness and accuracy of task execution.

[0065] In practical applications, in different application scenarios, the selected tasks executed on the first user may be different. For example: In the commodity marketing scenario, the selected tasks executed on the first user may be, for example: delivering commodity advertisements to the first user, issuing commodity coupons to the first user, sending commodity discount information to the first user, etc. In the credit scenario, the selected tasks executed on the first user may be, for example: delivering credit product advertisements to the first user, sending credit limit increase information to the first user, sending borrowing interest rate reduction information to the first user, etc.

[0066] Figure 2 In the method in [description], after obtaining the sequence operation information related to the target resource of the first user in the first time period, inputting the sequence operation information, the time interval in which the acquisition time of the sequence operation information is located, and m candidate tasks into the neural network model, the evaluation results of these m candidate tasks output by the neural network model can be obtained, and the candidate task with the highest probability that the first user obtains the target resource is selected as the selected task, and the selected task is executed on the first user within the time interval. Among them, the acquisition time of the sequence operation information can correspond to the current time. Thus, when the neural network model performs task inference, it needs to consider the time interval in which the current time is located; and, when the neural network model performs task inference, the collection time of the user's sequence operation information on which it is based is not far from the current time; furthermore, after the neural network model performs task inference, it needs to execute the selected task with the highest probability selected based on the inference result of the model within the time interval in which the current time is located. To sum up, on the one hand, by timely referring to the user's sequence operation information that is not far from the current time, the timeliness of the inference result output by the model itself is relatively high. In other words, it is ensured that the inference result output by the model has high timeliness; on the other hand, by referring to the time interval in which the current time is located when obtaining the inference result, and timely executing the selected task selected based on the inference result of the model within the time interval in which the current time is located, the high timeliness of task execution is guaranteed.

[0067] In addition, Figure 2 the method in can, by perceiving the user's sequential operation information in real time and making real-time decisions, timely capture the user's characteristics, and through deep learning algorithms, mine the user's needs in real time, so that the selected task determined based on the neural network model can meet the user's current personalized needs, which is beneficial to improving the accuracy of task execution.

[0068] Based on Figure 2 the method in , the embodiments of this specification also provide some specific implementation schemes of this method, which will be described below.

[0069] Optionally, Figure 2 in the method in , the operation information may include operation node information and node residence duration.

[0070] In the embodiments of this specification, each operation information in the sequential operation information may specifically include operation node information and the residence duration at this node. Among them, the operation node information can be obtained through the pre-arranged data points in the operation page. The operation node information can be, for example: information that the user browses a certain page, information that the user clicks a certain control, information that the user purchases a certain product, information that the user signs a certain agreement, etc. The node residence duration may refer to the time that the user stays at this node, which can reflect the user's attention to this node. For example: if the user spends 20 seconds browsing page A, then the node residence duration of this node is 20 seconds. By obtaining the user's operation node information and node residence duration, the user's operation can be more accurately characterized, so that the determined marketing task can be more accurate and the marketing effect can be better.

[0071] In practical applications, since it is necessary to determine the time partition where the acquisition moment of the sequential operation information is located, it is necessary to pre-divide each time partition.

[0072] Based on this, Figure 2 in the method in , before step 204, when inputting the sequential operation information, the time interval where the acquisition moment of the sequential operation information is located, and m candidate tasks into the neural network model to obtain the evaluation results of the m candidate tasks output by the neural network model, it may further include:

[0073] Obtaining the acquisition moment information of the sequential operation information;

[0074] According to the preset time partition rule, determining the time partition where the acquisition moment of the sequential operation information is located.

[0075] In the embodiments of this specification, the preset time partition rule may be to partition time intervals every 15 minutes, or every hour, or by morning / afternoon / evening (for example, it can be defined that the time interval from 8:00 to 12:00 is the morning time partition, the time interval from 12:00 to 18:00 is the afternoon time partition, and the time interval from 18:00 to 24:00 is the evening time partition), or it can also be to partition time intervals according to the peak hours of user Internet access (for example, define the time interval from 18:00 to 22:00 in a day as the peak hours of user Internet access, and the remaining time intervals as the non-peak hours of user Internet access). There is no specific limitation on this.

[0076] In practical applications, to avoid disturbing users at night, it is also possible not to partition the time interval for the night period. Correspondingly, it is also possible to avoid performing the selected tasks on users during the night period. For example, considering that users may need to rest between 20:00 and 8:00 the next day, it is also possible not to partition the time interval for this period, and at the same time, avoid performing the selected tasks on users during this period. Or, considering special business requirements, it is also possible to partition the time interval for the night period. There is no specific limitation on this.

[0077] In practical applications, it is also possible to obtain the geographical location information of the user and input the geographical location information of the user into the neural network model as well, so that the neural network model can also consider the geographical location information of the user when making predictions, improving the accuracy of the predictions.

[0078] Based on this, Figure 2 In the method in, in step 204, before inputting the sequence operation information, the time interval in which the acquisition time of the sequence operation information is located, and m candidate tasks into the neural network model to obtain the evaluation results of the m candidate tasks output by the neural network model, it may further include:

[0079] Obtain the geographical location information of the first user;

[0080] Correspondingly, the step of inputting the sequence operation information, the time interval in which the acquisition time of the sequence operation information is located, and m candidate tasks into the neural network model to obtain the evaluation results of the m candidate tasks output by the neural network model may specifically include:

[0081] Input the sequence operation information, the geographical location information, the time interval in which the acquisition time of the sequence operation information is located, and m candidate tasks into the neural network model to obtain the evaluation results of the m candidate tasks output by the neural network model.

[0082] In the embodiments of this specification, the terminal device of the first user can obtain the geographical location information of the first user in real time. After obtaining the user's authorization, it can report the geographical location information of the first user to the task execution platform in real time. Alternatively, after obtaining the user's authorization, the terminal device of the first user can also report the geographical location information of the first user to the task execution platform once every preset time period. Or, after obtaining the user's authorization, the task execution platform can also obtain the geographical location information of the first user while obtaining the sequence operation information related to the target resource of the first user within the first time period from the terminal device of the first user. No specific limitation is made thereto.

[0083] In the embodiments of this specification, the geographical location information of the first user is also input into the neural network model, so that the neural network model can also refer to the geographical location information of the first user during the calculation process, thereby being able to refer to the environment where the user is located, improving the accuracy of the evaluation result output by the model, and further being beneficial to improving the decision-making accuracy of the marketing task.

[0084] Optionally, the step of inputting the sequence operation information, the time interval in which the acquisition time of the sequence operation information is located, and m candidate tasks into the neural network model to obtain the evaluation results of the m candidate tasks output by the neural network model may specifically include:

[0085] Based on the geographical location information, determine the location type information where the first user is located;

[0086] Input the sequence operation information, the location type information, the time interval in which the acquisition time of the sequence operation information is located, and m candidate tasks into the neural network model to obtain the evaluation results of the m candidate tasks output by the neural network model.

[0087] In the embodiments of this specification, the geographical location information may be longitude and latitude information or other information that can represent an accurate coordinate position. The location type information may be, for example: shopping mall, residential community, school, restaurant, gym, office building, suburbs, highway, etc. Thus, the location type information where the user is located can be determined based on the geographical location information of the user. For example: after obtaining the longitude and latitude coordinate data of user A, it can be determined in combination with the electronic map data that user A is located within school B.

[0088] In the embodiments of this specification, when inputting the user's location information into the neural network model, the location type information where the user is located can be determined first based on the user's precise geographical location information, and then the location type information where the user is located is input into the neural network model. Since the location type information where the user is located is clearer in type and has fewer type categories compared to the user's geographical location information, using the location type information as a feature instead of using precise coordinate location information can reduce the variable space, lower the decision-making difficulty, and improve the decision-making efficiency while ensuring accuracy.

[0089] In practical applications, the user feature information of the user can also be obtained and input into the neural network model so that the neural network model can also consider the user feature information when making predictions.

[0090] Based on this, Figure 2 In the method in, in step 204, before inputting the sequence operation information, the time interval where the acquisition time of the sequence operation information is located, and m candidate tasks into the neural network model to obtain the evaluation results of the m candidate tasks output by the neural network model, it can further include:

[0091] Obtain the user feature information of the first user;

[0092] Correspondingly, the inputting the sequence operation information, the time interval where the acquisition time of the sequence operation information is located, and m candidate tasks into the neural network model to obtain the evaluation results of the m candidate tasks output by the neural network model can specifically include:

[0093] Input the sequence operation information, the geographical location information, the user feature information, the time interval where the acquisition time of the sequence operation information is located, and m candidate tasks into the neural network model to obtain the evaluation results of the m candidate tasks output by the neural network model.

[0094] In the embodiments of this specification, the user feature information may include the features of the user himself, such as: the user's credit limit, the user's credit amount, the number of visits by the user in the past 7 days, the user's occupation, the user's education level, etc. In practical applications, the user feature information can also be divided according to the relevance to the business. For example: the user feature information irrelevant to the business may include the user's occupation, the user's education level, etc.; the user feature information relevant to the business may include the user's credit limit, the user's credit amount, the number of visits by the user in the past 7 days, etc.

[0095] In practical applications, the terminal device of the first user can obtain the user characteristic information of the first user in real time. After obtaining the user's authorization, the terminal device of the first user can report the user characteristic information of the first user to the task execution platform in real time. Or, after obtaining the user's authorization, the terminal device of the first user can also report the user characteristic information of the first user to the task execution platform once every preset period. Or, after obtaining the user's authorization, the task execution platform can also obtain the user characteristic information of the first user while obtaining the sequence operation information related to the target resource of the first user in the first time period from the terminal device of the first user. There is no specific limitation on this.

[0096] In the embodiments of this specification, the user characteristic information of the first user is also input into the neural network model, so that the neural network model can also refer to the user characteristic information of the first user during the calculation process, thereby being able to refer to the user's own characteristics, which is beneficial to improving the accuracy of the evaluation results output by the model, and further beneficial to improving the decision-making accuracy of the marketing task, and can achieve personalized marketing.

[0097] Optionally, Figure 2 In the method in, in step 204, before inputting the sequence operation information, the time interval where the acquisition time of the sequence operation information is located, and m candidate tasks into the neural network model to obtain the evaluation results of the m candidate tasks output by the neural network model, it may further include:

[0098] Judge whether the first user meets the population screening conditions for obtaining the target resource to obtain a population judgment result;

[0099] Correspondingly, the inputting the sequence operation information, the time interval where the acquisition time of the sequence operation information is located, and m candidate tasks into the neural network model to obtain the evaluation results of the m candidate tasks output by the neural network model may specifically include:

[0100] If the population judgment result indicates that the first user meets the population screening conditions for obtaining the target resource, then input the sequence operation information, the time interval where the acquisition time of the sequence operation information is located, and m candidate tasks into the neural network model to obtain the evaluation results of the m candidate tasks output by the neural network model.

[0101] In the embodiments of this specification, the population screening conditions for obtaining the target resource may be conditions formulated by the provider or marketer of the target resource for screening the population. For example: If the population screening condition for obtaining the target resource is that "the user needs to be at least 18 years old", then users under 18 years old cannot meet the population screening conditions for obtaining the target resource.

[0102] In practical applications, instead of using the model to perform calculations for all users and executing the selected task for all users, the population is screened, and the selected task is only executed for users who meet the population screening criteria for obtaining the target resource. This can avoid disturbing users without business needs and can also save the model's computing resources and task execution resources. Taking a specific marketing scenario as an example, before pushing marketing information to users, it is usually necessary to screen the marketability of the marketing targets to avoid marketing users who do not meet the marketing conditions, which can not only avoid customer complaints but also avoid waste of marketing resources.

[0103] Optionally, the judgment of whether the first user meets the population screening criteria for obtaining the target resource may specifically include:

[0104] Judging whether the first user belongs to the pre-selected target population.

[0105] In the embodiments of this specification, the provider or marketer of the target resource may pre-select a part of the target population and only perform the determination and execution work of the selected task for this part of the target population. Specifically, the target population can be selected by means of a whitelist, and the target population that meets the conditions is added to the whitelist. Alternatively, other methods can also be used to pre-select the target population, and no specific limitation is made here.

[0106] In practical applications, by pre-selecting the target population, it is only necessary to compare whether the user is the target population, which can effectively improve the population screening efficiency.

[0107] In practical applications, the population can also be screened according to the business characteristic information of the users.

[0108] Based on this, the judgment of whether the first user meets the population screening criteria for obtaining the target resource may specifically include:

[0109] Obtain the business characteristic information of the first user;

[0110] According to the business characteristic information, judge whether the first user meets the preset target population conditions.

[0111] In the embodiments of this specification, the business characteristic information may be user characteristic information related to the business. In different business scenarios, the types of information included in the user's business characteristic information are different. For example, in the credit business scenario, the business characteristic information of the user may include, but is not limited to: the user's credit limit, the user's credit amount, the number of visits of the user, etc.

[0112] In the embodiments of this specification, according to the service characteristic information, it is determined whether the first user meets the preset target population condition. Specifically, the service characteristic information can be directly compared with the target population condition. For example: If the preset target population condition is that "the number of visits of the user in the recent seven days is not less than 3 times", and if the service characteristic information of the first user indicates that the number of visits of the first user in the recent seven days is 5 times, then the first user meets the preset target population condition; if the service characteristic information of the first user indicates that the number of visits of the first user in the recent seven days is 2 times, then the first user does not meet the preset target population condition.

[0113] In practical applications, since the service characteristic information of users usually changes dynamically, therefore, the current service characteristic information of the first user can be obtained in real time, and based on the service characteristic information, it is determined whether the first user currently meets the preset target population condition, which is beneficial to ensuring the timeliness and accuracy of population screening.

[0114] In practical applications, before obtaining the sequential operation information, geographical location information, user characteristic information, etc. of the user, the task execution platform can also first determine whether the user meets the population screening condition, and only obtain the relevant information of the users who meet the population screening condition, so as to reduce the acquisition of information of irrelevant populations, which is beneficial to improving the information acquisition efficiency and also beneficial to reducing the consumption of information transmission resources.

[0115] In practical applications, since the data points are pre-set, usually all users' information can be directly obtained, and then qualification screening is performed based on the obtained users' information. In this way, the existing data points can be reused without the need to specifically perform population screening and set data points in advance. This solution is beneficial to reducing the upfront resource investment.

[0116] Optionally, the candidate tasks may include first-type candidate tasks or second-type candidate tasks; the first-type candidate tasks are used to represent performing a resource recommendation operation for the target resource to the user; the second-type candidate tasks are used to represent not performing a resource recommendation operation for the target resource to the user.

[0117] In the embodiments of this specification, different first-type candidate tasks may represent different resource recommendation operations for the target resource to the user. For example: In the commodity marketing scenario, the first-type candidate tasks may be, for example: displaying commodity advertisements to the user, issuing commodity coupons to the user, sending commodity discount information to the user, etc. In the credit scenario, the first-type candidate tasks may be, for example: displaying credit product advertisements to the user, sending information about increased credit limits to the user, sending information about reduced borrowing interest rates to the user, etc.

[0118] In the embodiments of this specification, not performing the resource recommendation operation for the target resource can also be a candidate task, corresponding to the second type of candidate task. In practical applications, if the selected task is the first type of candidate task, the resource recommendation operation represented by the selected task can be performed on the first user within the time interval; if the selected task is the second type of selected task, no resource recommendation operation needs to be performed on the user.

[0119] Optionally, Figure 2 In the method in, in step 204, inputting the sequence operation information, the time interval in which the acquisition moment of the sequence operation information is located, and m candidate tasks into a neural network model to obtain the evaluation results of the m candidate tasks output by the neural network model may specifically include:

[0120] Based on the sequence operation information, the time interval in which the acquisition moment of the sequence operation information is located, and m candidate tasks, m candidate solution features are obtained;

[0121] Input the m candidate solution features into a neural network model for processing sequence data to obtain the evaluation results of each candidate task among the m candidate tasks output by the neural network model.

[0122] In the embodiments of this specification, there are differences between candidate solutions and candidate tasks. A candidate solution can be used to represent that when a first user performs the specified sequence operation information, a specified candidate task is performed within the specified time interval. That is, the sequence operation information, the time interval in which the acquisition moment of the sequence operation information is located, and 1 candidate task can be combined into 1 candidate solution.

[0123] In practical applications, when forming a candidate solution, at least one of the user's geographical location information and the user's user feature information can also be added to improve the accuracy of the model output result. For example: the user's sequence operation information, the time interval in which the acquisition moment of the sequence operation information is located, the user's geographical location information, the user's user feature information, and 1 candidate task can be combined into 1 candidate solution. Thus, m candidate tasks can form m combinations to obtain m candidate solutions.

[0124] In practical applications, sequence features for input into a neural network model can be generated based on the sequence operation information; discrete features corresponding to each of the m candidate tasks can be generated based on the user's geographical location information, the user's user feature information, the time interval in which the acquisition time of the sequence operation information is located, and the m candidate tasks; then, the sequence features and the discrete features can be input into a neural network model for processing sequence data. By converting information, tasks, and time intervals into data features that can be processed by the model, the normal processing of the model can be ensured, which is beneficial to improving the model processing efficiency.

[0125] In practical applications, since executing tasks for users consumes resources, therefore, considering the marketing cost, after determining the selected task, it can first be determined whether the preset marketing cost has been exceeded, and the selected task can be executed only when the cost requirement is met.

[0126] Based on this, Figure 2 In the method in, in step 208, before executing the selected task on the first user within the time interval, it may further include:

[0127] Obtain the resource consumption amounts of each candidate task;

[0128] According to the resource consumption amounts and the selected tasks of multiple users including the first user, determine the resource consumption indication amounts of the multiple users;

[0129] Judge whether the resource consumption indication amount is less than or equal to a preset resource threshold to obtain a resource consumption judgment result;

[0130] Correspondingly, the execution of the selected task on the first user within the time interval may specifically include:

[0131] If the resource consumption judgment result indicates that the resource consumption indication amount is less than or equal to the preset resource threshold, then execute the selected task on the first user within the time interval.

[0132] In the embodiments of this specification, the resource consumption amount may include but is not limited to: capital consumption amount, human resource consumption amount, marketing resource consumption amount, computing resource consumption amount, etc. Executing each candidate task will have corresponding resource consumption. For each candidate task, the corresponding resource consumption amount can be calculated according to the content of the candidate task, or the resource consumption amounts of each candidate task can also be determined by other means, and no specific limitation is made thereto.

[0133] In the embodiments of this specification, the resource consumption indication quantities of multiple users may include the total resource consumption of multiple users, or the average resource consumption of each user, etc. The preset resource threshold can be set and adjusted according to actual needs, and no specific limitation is imposed on the specific value of the preset resource threshold.

[0134] In practical applications, based on the resource consumption indication quantities of multiple users including the first user, to determine whether to execute a selected task for the first user within the time interval, overall resource consumption control can be performed based on multiple users. If executing the selected task for the first user will cause the resource consumption to exceed the preset threshold, then it is prohibited to execute the selected task for the first user within the time interval. Thereby, it is possible to avoid resource consumption exceeding the budget, which is beneficial for reasonable resource management and control.

[0135] Figure 3 It is a flowchart of a method for population screening and task execution provided by the embodiments of this specification. As Figure 3 shown, the method for population screening and task execution may specifically include the following steps:

[0136] Step 302: Obtain the sequence operation information, geographical location information, and user characteristic information of the first user.

[0137] Step 304: Determine whether the first user meets the population screening conditions for obtaining the target resources. If so, jump to step 306; if not, end.

[0138] Step 306: Input the sequence operation information, geographical location information, user characteristic information, the time interval where the acquisition moment of the sequence operation information is located, and m candidate tasks of the first user into the neural network model.

[0139] Step 308: Obtain the evaluation results of the m candidate tasks output by the neural network model.

[0140] Step 310: Select the task with the highest probability from the m candidate tasks as the selected task according to the evaluation results.

[0141] Step 312: Determine whether the resource consumption indication quantities of multiple users including the first user are less than or equal to the preset resource threshold. If so, jump to step 314; if not, end.

[0142] Step 314: Execute the selected task for the first user within the time interval where the acquisition moment of the sequence operation information is located.

[0143] In practical applications, before using a neural network model to predict the probability that a user obtains a target resource after performing the candidate task on the user within the time interval based on the user's sequence operation information, the time interval in which the acquisition time of the sequence operation information is located, and the candidate task, it is also necessary to train the neural network model using samples.

[0144] Based on this, Figure 2 In the method in, in step 204, before inputting the sequence operation information, the time interval in which the acquisition time of the sequence operation information is located, and m candidate tasks into the neural network model to obtain the evaluation results of the m candidate tasks output by the neural network model, it may further include:

[0145] Obtain a sample plan carrying a sample label; the sample plan is used to represent that after a second user performs a sample sequence operation, a sample task is performed on the second user within a sample time interval; the sample task includes a first type of sample task or a second type of sample task; the first type of sample task is used to represent that a resource recommendation operation for a sample resource is performed on the second user; the second type of sample task is used to represent that a resource recommendation operation for the sample resource is not performed on the second user; the sample label includes a first sample label or a second sample label, the first sample label is used to represent that after performing the sample plan on the second user, the second user obtains the sample resource; the second sample label is used to represent that after performing the sample plan on the second user, the second user does not obtain the sample resource;

[0146] Input the features corresponding to the sample plan into the neural network model to be trained, and obtain the prediction results output by the neural network model to be trained;

[0147] According to the prediction results and the sample labels, adjust the parameters of the neural network model to be trained to obtain a trained neural network model.

[0148] In the embodiments of this specification, the sample sequence operation may be several operation information sorted by occurrence time of the second user within a second time period, and the time interval between the end time of the second time period and the acquisition time of the sample sequence operation is less than or equal to a first duration threshold. The sample time interval is the time interval in which the acquisition time of the sample sequence operation is located. And the sample task is performed on the second user within the sample time interval.

[0149] In the embodiments of this specification, the features corresponding to the sample solution are input into the neural network model to be trained. Specifically, the feature vectors corresponding to the sample sequence operations, the feature vectors corresponding to the sample time intervals, and the feature vectors corresponding to the sample tasks can all be input into the neural network model to be trained. Alternatively, it can also be that after fusing the feature vectors corresponding to the sample sequence operations, the feature vectors corresponding to the sample time intervals, and the feature vectors corresponding to the sample tasks, the fused features are then input into the neural network model to be trained, and no specific limitation is made in this regard.

[0150] In practical applications, the sample solution may further include: the sample position information of the second user at the acquisition moment of the sample sequence operation. More specifically, during training, based on the sample position information, the sample position type information of the second user can also be determined, and the features (vectors) of the sample position type information are also input into the neural network model to be trained. The sample position type information can be such as: shopping mall, residential community, school, restaurant, gym, office building, suburb, road, etc. By also inputting the sample position information or the sample position type information of the second user into the neural network model to be trained, the neural network model to be trained can also learn the characteristics related to the user's position information during the training process, so that the accuracy of the evaluation result output by the model can be improved with reference to the environment where the user is located.

[0151] Furthermore, the sample solution may further include: the user feature information of the second user. More specifically, during training, the features (vectors) of the user feature information can also be input into the neural network model to be trained. By also inputting the user feature information of the second user into the neural network model to be trained, the neural network model to be trained can also learn the characteristics related to the user's feature information during the training process, so that the accuracy of the evaluation result output by the model can be improved with reference to the user's own characteristics.

[0152] In practical applications, the termination conditions for the process of training the neural network model using samples may include reaching the convergence condition or reaching the preset iteration number threshold. Among them, the convergence condition may include that the loss value of the model loss function reaches the preset threshold. For example: If the preset iteration number threshold is 500 times, then the neural network model is trained using samples, and after 500 iterations, the training ends. Another example: The set convergence condition is that the loss value of the model loss function is less than 0.01, then the neural network model is trained using samples, and when the loss value of the model loss function is less than 0.01, the training ends.

[0153] In practical applications, the training samples can also be divided into a training set and a test set. After training the model using the samples in the training set, the performance of the trained model (such as the accuracy, precision, recall, etc. of the model) can be tested using the samples in the test set. After the performance of the model meets the standard, the training can be ended. If it is found that the performance of the model does not meet the standard after testing, the training parameters of the model can be adjusted and then the training can continue, or the model can be further trained using new samples.

[0154] Figure 4 It is a flowchart of a neural network model training method provided by an embodiment of this specification. As Figure 4 shown, the neural network model training method can specifically include the following steps:

[0155] Step 402: Obtain a sample solution carrying sample labels.

[0156] Step 404: Divide the sample solution into a training set and a test set.

[0157] Step 406: Input the features corresponding to the sample solution in the training set into the neural network model.

[0158] Step 408: The model performs forward propagation, calculates the loss value, and performs backward propagation.

[0159] Step 410: Update the model weights.

[0160] Step 412: Determine whether the model training termination condition is met. If the model training termination condition is met, jump to step 414; if the model training termination condition is not met, jump back to step 406. Among them, the model training termination condition can include: the loss value of the model loss function reaches a preset threshold, or the number of iterations reaches a preset iteration number threshold.

[0161] Step 414: Evaluate the accuracy of the neural network model using the sample solution in the test set.

[0162] Step 416: Determine whether the accuracy of the model meets the standard. If the accuracy of the model meets the standard, jump to step 418; if the accuracy of the model does not meet the standard, jump to step 420.

[0163] Step 418: Obtain the trained neural network model.

[0164] Step 420: Adjust the model training parameters or update the samples, and jump back to step 406.

[0165] Optionally, before obtaining the sample solution carrying sample labels, it can further include:

[0166] Execute a sample task for the second user;

[0167] Associate and store the sample task, the sample time interval to which the execution time of the sample task belongs, and the sample sequence operation of the second user before executing the sample task as a sample solution;

[0168] Obtain the user feedback behavior information of the second user in response to the sample task;

[0169] Determine a sample label corresponding to the sample task according to the user feedback behavior information;

[0170] Mark the sample solution with the sample label.

[0171] In the embodiments of this specification, the second user may be a user who meets the population screening conditions for obtaining the target resource. In practical applications, multiple second users may be selected, and multiple sample solutions may be constructed based on the behavior information feedback by the multiple second users in response to the sample task.

[0172] In practical applications, after performing the sample task on the second user, the user feedback behavior information of the second user in response to the sample task may be collected within a preset time period. Among them, the preset time period can be set and adjusted according to actual needs. For example: after performing the sample task on the second user, collect the user feedback behavior information of the second user within 24 hours. Another example: after performing the sample task on the second user, collect the user feedback behavior information of the second user within 3 hours. No specific limitation is made in this regard.

[0173] In practical applications, determine a sample label corresponding to the sample task according to the user feedback behavior information of the second user. Specifically, if the user feedback behavior information of the second user indicates that the second user has obtained the target resource, the sample label can be marked as "1"; if the user feedback behavior information of the second user indicates that the second user has not obtained the target resource, the sample label can be marked as "0". Or, in addition to using "0" and "1" as sample labels, other sample labels can also be used, and no specific limitation is made in this regard.

[0174] Exemplarily, taking a specific commodity marketing scenario as an example, the sample task executed by the task execution platform to user A is "send a commodity coupon for commodity a", and the user feedback behavior information of user A obtained later is "user A purchased commodity a", then the corresponding sample label can be determined as "1".

[0175] In practical applications, the sample solution may further include the sample position information at the acquisition moment of the sample sequence operation of the second user and the user feature information of the second user. That is: the sample task, the sample time interval to which the execution moment of the sample task belongs, the sample sequence operation of the second user before executing the sample task, the sample position information at the acquisition moment of the sample sequence operation of the second user, and the user feature information of the second user can be associated and stored as a sample solution.

[0176] Figure 5 It is a flowchart of a method for constructing a sample solution provided by an embodiment of this specification. As Figure 5 shown, the method for constructing a sample solution may specifically include the following steps:

[0177] Step 502: Select a second user who meets the population screening conditions for obtaining the target resource.

[0178] Step 504: Execute a sample task for the second user.

[0179] Step 506: Associate and store the sample task, the sample time interval to which the execution moment of the sample task belongs, and the sample sequence operation of the second user before executing the sample task as a sample solution.

[0180] Step 508: Obtain the user feedback behavior information of the second user in response to the sample task.

[0181] Step 510: Determine the sample label corresponding to the sample task according to the user feedback behavior information.

[0182] Step 512: Mark the sample solution with the sample label to obtain a sample solution carrying the sample label.

[0183] Based on the same idea, an embodiment of this specification also provides a device corresponding to the above method. Figure 6 It is a schematic structural diagram of a task execution device corresponding to Figure 2 provided by an embodiment of this specification. As Figure 6 shown, the device may include:

[0184] An information acquisition module 602, configured to acquire the sequence operation information of the first user related to the target resource within the first time period; the time interval between the end moment of the first time period and the acquisition moment of the sequence operation information is less than or equal to the first duration threshold; the first duration threshold is at the hour level or the minute level; the sequence operation information includes a number of operation information sorted according to the occurrence time.

[0185] A model inference module 604, configured to input the sequence operation information, the time interval in which the acquisition moment of the sequence operation information is located, and m candidate tasks into a neural network model, and obtain an evaluation result of the m candidate tasks output by the neural network model; the time interval is at the hour level or the minute level; m is an integer greater than or equal to 2; the evaluation result is used to reflect the probability that the first user obtains the target resource after performing the candidate task on the first user.

[0186] A task selection module 606, configured to determine a selected task to be performed on the first user from the m candidate tasks according to the evaluation result; the probability that the first user corresponding to the selected task obtains the target resource is higher than that of other candidate tasks.

[0187] A task execution module 608, configured to perform the selected task on the first user within the time interval.

[0188] Based on Figure 6 the apparatus, some specific implementation manners of the apparatus are further provided in the embodiments of this specification, and are described below.

[0189] Optionally, the operation information includes operation node information and node residence duration.

[0190] Optionally, the apparatus may further include:

[0191] A moment information acquisition module, configured to acquire the acquisition moment information of the sequence operation information.

[0192] A time partition determination module, configured to determine the time partition in which the acquisition moment of the sequence operation information is located according to a preset time partition rule.

[0193] Optionally, the apparatus may further include:

[0194] A geographical location information acquisition module, configured to acquire the geographical location information of the first user.

[0195] Correspondingly, the model inference module 604 may specifically include:

[0196] A first model inference unit, configured to input the sequence operation information, the geographical location information, the time interval in which the acquisition moment of the sequence operation information is located, and m candidate tasks into a neural network model, and obtain an evaluation result of the m candidate tasks output by the neural network model.

[0197] Optionally, the first model inference unit may specifically include:

[0198] A location type information determination subunit, configured to determine the location type information of the first user based on the geographical location information.

[0199] A model inference subunit, configured to input the sequence operation information, the location type information, the time interval in which the acquisition moment of the sequence operation information is located, and m candidate tasks into a neural network model, and obtain an evaluation result of the m candidate tasks output by the neural network model.

[0200] Optionally, the apparatus may further include:

[0201] A user feature information acquisition module, configured to acquire the user feature information of the first user.

[0202] Correspondingly, the model inference module 604 may specifically include:

[0203] A second model inference unit, configured to input the sequence operation information, the geographical location information, the user feature information, the time interval in which the acquisition moment of the sequence operation information is located, and m candidate tasks into a neural network model, and obtain an evaluation result of the m candidate tasks output by the neural network model.

[0204] Optionally, the apparatus may further include:

[0205] A judgment module, configured to judge whether the first user meets the population screening condition for acquiring the target resource, and obtain a population judgment result.

[0206] Correspondingly, the model inference module 604 may specifically include:

[0207] A third model inference unit, configured to, if the population judgment result indicates that the first user meets the population screening condition for acquiring the target resource, input the sequence operation information, the time interval in which the acquisition moment of the sequence operation information is located, and m candidate tasks into a neural network model, and obtain an evaluation result of the m candidate tasks output by the neural network model.

[0208] Optionally, the judgment module may specifically include:

[0209] A first judgment unit, configured to judge whether the first user belongs to a pre-selected target population.

[0210] Optionally, the judgment module may specifically include:

[0211] A service feature information acquisition unit, configured to acquire the service feature information of the first user.

[0212] A second judgment unit, configured to judge whether the first user meets a preset target population condition according to the service feature information.

[0213] Optionally, in the device, the candidate tasks include first-type candidate tasks or second-type candidate tasks; the first-type candidate tasks are used to represent performing a resource recommendation operation for the target resource to the user; the second-type candidate tasks are used to represent not performing a resource recommendation operation for the target resource to the user.

[0214] Optionally, the model inference module 604 may specifically include:

[0215] A feature determination unit, configured to obtain m candidate solution features based on the sequence operation information, the time interval where the acquisition moment of the sequence operation information is located, and the m candidate tasks.

[0216] An input unit, configured to input the m candidate solution features into a neural network model for processing sequence data, and obtain an evaluation result of each candidate task in the m candidate tasks output by the neural network model.

[0217] Optionally, the device may further include:

[0218] A resource consumption amount acquisition module, configured to acquire the resource consumption amounts of the respective candidate tasks.

[0219] A resource consumption indication amount determination module, configured to determine the resource consumption indication amounts of the multiple users according to the resource consumption amounts and the selected tasks of the multiple users including the first user.

[0220] A resource consumption judgment module, configured to judge whether the resource consumption indication amount is less than or equal to a preset resource threshold, and obtain a resource consumption judgment result.

[0221] Correspondingly, the task execution module 608 may specifically include:

[0222] A task execution unit, configured to, if the resource consumption judgment result indicates that the resource consumption indication amount is less than or equal to the preset resource threshold, perform the selected task on the first user within the time interval.

[0223] Optionally, the device may further include:

[0224] A sample plan acquisition module, configured to acquire a sample plan carrying a sample label; the sample plan is used to represent that after a second user performs a sample sequence operation, a sample task is performed on the second user within a sample time interval; the sample task includes a first type of sample task or a second type of sample task; the first type of sample task is used to represent that a resource recommendation operation for sample resources is performed on the second user; the second type of sample task is used to represent that a resource recommendation operation for the sample resources is not performed on the second user; the sample label includes a first sample label or a second sample label, the first sample label is used to represent that after the sample plan is performed on the second user, the second user acquires the sample resources; the second sample label is used to represent that after the sample plan is performed on the second user, the second user does not acquire the sample resources.

[0225] A prediction result determination module, configured to input the features corresponding to the sample plan into a neural network model to be trained, and obtain a prediction result output by the neural network model to be trained.

[0226] A model parameter adjustment module, configured to adjust the parameters of the neural network model to be trained according to the prediction result and the sample label, and obtain a trained neural network model.

[0227] Optionally, the apparatus may further include:

[0228] A sample task execution module, configured to perform a sample task on a second user.

[0229] A sample plan storage module, configured to associate and store the sample task, the sample time interval to which the execution time of the sample task belongs, and the sample sequence operation of the second user before performing the sample task as a sample plan.

[0230] A user feedback behavior information acquisition module, configured to acquire user feedback behavior information of the second user in response to the sample task.

[0231] A sample label determination module, configured to determine a sample label corresponding to the sample task according to the user feedback behavior information.

[0232] A marking module, configured to mark the sample plan with the sample label.

[0233] It can be understood that the above-mentioned modules refer to computer programs or program segments for performing one or more specific functions. In addition, the distinction between the above-mentioned modules does not mean that the actual program codes must also be separated.

[0234] Based on the same idea, an embodiment of this specification also provides a device corresponding to the above method.

[0235] Figure 7 A structural schematic diagram of a task execution device provided in an embodiment of this specification corresponding to Figure 2 . As shown in Figure 7 , the device 700 may include: at least one processor 710; and a memory 730 communicatively connected to the at least one processor; wherein, the memory 730 stores instructions 720 executable by the at least one processor 710, and the instructions are executed by the at least one processor 710 to enable the at least one processor 710 to:

[0236] Obtain sequence operation information of a first user related to a target resource within a first time period; the time interval between the end time of the first time period and the acquisition time of the sequence operation information is less than or equal to a first duration threshold; the first duration threshold is at the hour level or the minute level; the sequence operation information includes a number of operation information sorted by occurrence time.

[0237] Input the sequence operation information, the time interval where the acquisition time of the sequence operation information is located, and m candidate tasks into a neural network model to obtain an evaluation result of the m candidate tasks output by the neural network model; the time interval is at the hour level or the minute level; m is an integer greater than or equal to 2; the evaluation result is used to reflect the probability that the first user obtains the target resource after the candidate task is executed for the first user.

[0238] Determine a selected task to be executed for the first user from the m candidate tasks according to the evaluation result; the probability that the first user corresponding to the selected task obtains the target resource is higher than that of other candidate tasks.

[0239] Execute the selected task for the first user within the time interval.

[0240] Each embodiment in this specification is described in a progressive manner. For the same or similar parts between each embodiment, reference can be made to each other. The key point of each embodiment is to illustrate the differences from other embodiments. In particular, for Figure 7 the device shown in , since it is basically similar to the method embodiment, the description is relatively simple. For related parts, reference can be made to the partial description of the method embodiment.

[0241] The foregoing describes particular embodiments of the present specification. Other embodiments are within the scope of the appended claims. In some cases, the acts or steps recited in the claims may be performed in a different order than in the embodiments and still achieve the desired result. Additionally, the processes depicted in the figures do not necessarily require the particular order shown or sequential order to achieve the desired result. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0242] In the 1990s, improvements to a technology could be clearly distinguished as either hardware improvements (e.g., improvements to circuit structures such as diodes, transistors, switches, etc.) or software improvements (improvements to method flows). However, with the development of technology, many method flow improvements today can be regarded as direct improvements to hardware circuit structures. Designers almost always obtain the corresponding hardware circuit structure by programming the improved method flow into the hardware circuit. Therefore, it cannot be said that an improvement to a method flow cannot be implemented using a hardware entity module. For example, a Programmable Logic Device (PLD) (such as a Field Programmable Gate Array (FPGA)) is an integrated circuit whose logic function is determined by the user programming the device. Designers can program themselves to "integrate" a digital system onto a single PLD, without having to ask a chip manufacturer to design and fabricate a dedicated integrated circuit chip. Moreover, today, instead of manually fabricating integrated circuit chips, this programming is mostly implemented using "logic compiler" software, which is similar to the software compilers used in program development and writing. The original code before compilation also has to be written in a specific programming language, which is called a Hardware Description Language (HDL). There is not just one type of HDL, but many, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. The most commonly used ones currently are VHDL (Very-High-Speed Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also be aware that by simply performing a little logical programming on the method flow using the above-mentioned several hardware description languages and programming it into an integrated circuit, it is easy to obtain the hardware circuit that implements the logical method flow.

[0243] The controller can be implemented in any suitable manner. For example, the controller can take the form of, for example, a microprocessor or a processor and a computer-readable medium storing computer-readable program code (such as software or firmware) executable by the (micro)processor, logic gates, switches, an application specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller. Examples of the controller include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320. The memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art also know that in addition to implementing the controller in the form of pure computer-readable program code, it is entirely possible to make the controller implement the same function in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, such a controller can be considered a hardware component, and the devices included therein for implementing various functions can also be regarded as the structures within the hardware component. Or even, the devices for implementing various functions can be regarded as either software modules for implementing the method or structures within the hardware component.

[0244] The systems, devices, modules, or units illustrated in the above embodiments can be specifically implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer can be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0245] For the convenience of description, when describing the above devices, they are described separately as various units according to their functions. Of course, when implementing this specification, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0246] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program code.

[0247] The present invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each flow and / or block of the flowchart illustrations and / or block diagrams, and combinations of flows and / or blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to the processors of a general purpose computer, special purpose computer, embedded processor or other programmable data processing apparatus to produce a machine, such that the instructions executed by the processors of the computer or other programmable data processing apparatus create means for implementing the functions specified in the flow Figure 1 one or more flows and / or blocks Figure 1 or means for implementing the functions specified in a block or blocks.

[0248] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means that implement the functions specified in the flow Figure 1 one or more flows and / or blocks Figure 1 or means for implementing the functions specified in a block or blocks.

[0249] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in the flow Figure 1 one or more flows and / or blocks Figure 1 or means for implementing the functions specified in a block or blocks.

[0250] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0251] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM) and / or non-volatile memory, such as read only memory (ROM) or flash memory (flash RAM). The memory is an example of computer-readable media.

[0252] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules 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 disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0253] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0254] This specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. This specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0255] The above description is only an embodiment of the present specification and is not intended to limit the present specification. For those skilled in the art, the present specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present specification shall be included in the scope of the claims of the present specification.

Claims

1. A task execution method, comprising: Acquire sequence operation information related to the target resource by the first user in a first time period; The time interval between the end time of the first time period and the acquisition time of the sequence operation information is less than or equal to a first time threshold; the first time threshold is in hours or minutes; the sequence operation information includes a plurality of operation information sorted by occurrence time; The sequence operation information, the time interval at which the acquisition time of the sequence operation information is located, and m candidate tasks are input into a neural network model to obtain evaluation results of the m candidate tasks output by the neural network model; the time interval is at the hour level or the minute level; m is an integer greater than or equal to 2; the evaluation result is used to reflect the probability of the first user acquiring the target resource after the candidate task is executed on the first user; According to the evaluation result, a selected task for executing on the first user is determined from the m candidate tasks; the probability of the first user acquiring the target resource corresponding to the selected task is higher than that of other candidate tasks; The selected task is performed on the first user during the time interval.

2. The method according to claim 1, wherein the operation information includes operation node information and node residence time.

3. The method according to claim 1, before inputting the sequence operation information, the time interval at which the acquisition time of the sequence operation information is located, and the m candidate tasks into a neural network model and obtaining the evaluation results of the m candidate tasks output by the neural network model, further comprising: Acquire acquisition time information of the sequence operation information; According to a preset time partition rule, the time partition where the acquisition time of the sequence operation information is located is determined.

4. The method according to claim 1, before inputting the sequence operation information, the time interval of the acquisition time of the sequence operation information, and the m candidate tasks into a neural network model and obtaining the evaluation results of the m candidate tasks output by the neural network model, further comprising: Acquire geographic location information of the first user; The step of inputting the sequence operation information, the time interval of the acquisition moment of the sequence operation information, and the m candidate tasks into a neural network model, and obtaining the evaluation results of the m candidate tasks output by the neural network model, specifically includes: The sequence operation information, the geographic location information, the time interval of the acquisition moment of the sequence operation information, and m candidate tasks are input into a neural network model to obtain evaluation results of the m candidate tasks output by the neural network model.

5. The method according to claim 4, wherein the sequence operation information, the geographic location information, the time interval at which the sequence operation information is obtained, and the m candidate tasks are input into a neural network model to obtain the evaluation results of the m candidate tasks output by the neural network model, specifically comprising: Based on the geographic location information, determine the location type information of the first user; The sequence operation information, the position type information, the time interval of the acquisition moment of the sequence operation information, and m candidate tasks are input into a neural network model to obtain evaluation results of the m candidate tasks output by the neural network model.

6. The method according to claim 4, before inputting the sequence operation information, the time interval at which the acquisition time of the sequence operation information is located, and the m candidate tasks into a neural network model and obtaining the evaluation results of the m candidate tasks output by the neural network model, further comprising: Acquire user characteristic information of the first user; The step of inputting the sequence operation information, the time interval of the acquisition moment of the sequence operation information, and the m candidate tasks into a neural network model, and obtaining the evaluation results of the m candidate tasks output by the neural network model, specifically includes: The sequence operation information, the geographic location information, the user characteristic information, the time interval of the acquisition moment of the sequence operation information, and m candidate tasks are input into a neural network model to obtain evaluation results of the m candidate tasks output by the neural network model.

7. The method according to claim 1, before inputting the sequence operation information, the time interval of the acquisition time of the sequence operation information, and the m candidate tasks into a neural network model and obtaining the evaluation results of the m candidate tasks output by the neural network model, further comprising: Determine whether the first user meets the population screening condition for obtaining the target resource, and obtain a population determination result; The step of inputting the sequence operation information, the time interval of the acquisition moment of the sequence operation information, and the m candidate tasks into a neural network model, and obtaining the evaluation results of the m candidate tasks output by the neural network model, specifically includes: If the crowd judgment result indicates that the first user meets the crowd screening conditions for obtaining the target resource, the sequence operation information, the time interval of the acquisition time of the sequence operation information, and m candidate tasks are input into the neural network model to obtain the evaluation results of the m candidate tasks output by the neural network model.

8. The method according to claim 7, wherein determining whether the first user meets a crowd screening condition for obtaining the target resource specifically comprises: Determine whether the first user belongs to the pre-selected target group.

9. The method according to claim 7, wherein determining whether the first user meets a crowd screening condition for obtaining the target resource specifically comprises: Acquiring service feature information of the first user; According to the service characteristic information, it is determined whether the first user meets a preset target population condition.

10. The method of claim 1, wherein: The candidate tasks include first-category candidate tasks or second-category candidate tasks; the first-category candidate tasks are used to indicate that a resource recommendation operation is performed on the target resource to the user; the second-category candidate tasks are used to indicate that there is no need to perform a resource recommendation operation on the target resource to the user.

11. The method according to claim 1, wherein the inputting the sequence operation information, the time interval of the acquisition time of the sequence operation information, and m candidate tasks into a neural network model to obtain the evaluation results of the m candidate tasks output by the neural network model specifically comprises: Based on the sequence operation information, the time interval of the acquisition time of the sequence operation information, and the m candidate tasks, m candidate solution features are obtained; The features of the m candidate solutions are input into a neural network model for processing sequence data, and an evaluation result of each of the m candidate tasks output by the neural network model is obtained.

12. The method of claim 1, before executing the selected task on the first user within the time interval, further comprising: Get the resource consumption of each candidate task; Determining resource consumption indicator amounts of the multiple users according to the resource consumption amount and the selected tasks of the multiple users including the first user; Determine whether the resource consumption indicator is less than or equal to a preset resource threshold, and obtain a resource consumption determination result; The performing the selected task on the first user within the time interval specifically includes: If the resource consumption determination result indicates that the resource consumption indicator amount is less than or equal to a preset resource threshold, the selected task is executed for the first user within the time interval.

13. The method of claim 1, before inputting the sequence operation information, the time interval of the acquisition time of the sequence operation information, and the m candidate tasks into a neural network model and obtaining the evaluation results of the m candidate tasks output by the neural network model, further comprising: A sample solution carrying a sample label is obtained; the sample solution is used to indicate that after the second user performs a sample sequence operation, a sample task is performed on the second user within a sample time interval; the sample task includes a first type of sample task or a second type of sample task; the first type of sample task is used to indicate that a resource recommendation operation for a sample resource is performed on the second user; the second type of sample task is used to indicate that a resource recommendation operation for the sample resource is not performed on the second user; the sample label includes a first sample label or a second sample label, and the first sample label is used to indicate that after the sample solution is performed on the second user, the second user obtains the sample resource; The second sample tag is used to indicate that after executing the sample solution for the second user, the second user does not obtain the sample resource; Inputting the features corresponding to the sample scheme into the neural network model to be trained, and obtaining the prediction results output by the neural network model to be trained; According to the prediction results and the sample labels, the parameters of the neural network model to be trained are adjusted to obtain a trained neural network model.

14. The method according to claim 13, before obtaining the sample solution carrying the sample tag, further comprising: performing the sample task to the second user; The sample task, the sample time interval to which the execution time of the sample task belongs, and the sample sequence operation of the second user before executing the sample task are associated and stored as a sample plan; Acquiring user feedback behavior information of the second user after responding to the sample task; Determining a sample label corresponding to the sample task according to the user feedback behavior information; The sample scheme is labeled using the sample tag.

15. A task execution device, comprising: An information acquisition module, used to acquire sequence operation information related to the target resource by the first user in the first time period; The time interval between the end time of the first time period and the acquisition time of the sequence operation information is less than or equal to a first time threshold; the first time threshold is in hours or minutes; the sequence operation information includes a plurality of operation information sorted by occurrence time; A model reasoning module, used for inputting the sequence operation information, the time interval of the acquisition time of the sequence operation information, and m candidate tasks into a neural network model, and obtaining an evaluation result of the m candidate tasks output by the neural network model; the time interval is at the hour level or the minute level; m is an integer greater than or equal to 2; the evaluation result is used to reflect the probability of the first user acquiring the target resource after the candidate task is executed on the first user; a task selection module, configured to determine, according to the evaluation result, a selected task for execution on the first user from the m candidate tasks; the probability of the first user acquiring the target resource corresponding to the selected task being higher than that of other candidate tasks; A task execution module is used to execute the selected task for the first user within the time interval.

16. A task execution device, comprising: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to: Acquire sequence operation information related to the target resource by the first user in a first time period; the time interval between the end time of the first time period and the acquisition time of the sequence operation information is less than or equal to a first time threshold; the first time threshold is in hours or minutes; the sequence operation information includes a plurality of operation information sorted by occurrence time; The sequence operation information, the time interval at which the acquisition time of the sequence operation information is located, and m candidate tasks are input into a neural network model to obtain evaluation results of the m candidate tasks output by the neural network model; the time interval is at the hour level or the minute level; m is an integer greater than or equal to 2; the evaluation result is used to reflect the probability of the first user acquiring the target resource after the candidate task is executed on the first user; According to the evaluation result, a selected task for executing on the first user is determined from the m candidate tasks; the probability of the first user acquiring the target resource corresponding to the selected task is higher than that of other candidate tasks; The selected task is performed on the first user during the time interval.

Images