WO2020098251A1 - User preference-based crowdsourced task pushing method and related device - Google Patents

Abstract

Disclosed by embodiments of the present application are a user preference-based crowdsourced task pushing method and a related device, wherein the method comprises: determining a first preset number of task response active users for a task of a target task type; acquiring a historical push data set of each task response active user respectively; establishing an acceptance probability logic regression model according to the historical push data set; estimating historical task acceptance probability according to active user data and historical task data by using the acceptance probability logic regression model; estimating the user likelihood of the task response active users according to the historical task acceptance probability and response result tag data; and determining a first high-preference target user from among the first preset number of task response active users according to the user likelihoods respectively corresponding to the first preset number of task response active users; and pushing a target push task of the target task type to the first high-preference target user. By employing the solution of the present application, the delivery efficiency of crowdsourced tasks may be improved.

Description

Crowdsourcing task pushing method and related device based on user preference

This application requires the priority of the Chinese patent application submitted to the China Patent Office on November 15, 2018, with the application number 2018113647111 and the application name "Crowdsourcing Task Push Method and Related Devices Based on User Preference", the entire content of which is cited Incorporated in this application.

Technical field

The present application relates to the field of computers, and in particular, to a crowdsourcing task pushing method and related device based on user preference.

Background technique

In recent years, more and more companies have begun to entrust certain technical tasks to external individuals or organizations through Internet channels. This emerging Internet-based open collaborative innovation model is called crowdsourcing. The crowdsourcing model generally includes three main bodies: the contractor, the crowdsourcing platform, and the contractor. The contractor generally publishes the crowdsourcing tasks on the crowdsourcing platform, and then the contractor undertakes the tasks from the crowdsourcing platform and completes them as agreed Task and get paid accordingly. Common crowdsourcing tasks include design, management consulting, scheme planning, film and television production, information collection, picture recognition, software development, etc. When crowdsourcing tasks are allocated, the crowdsourcing platform will widely distribute the task requirements of the issuer, and at the same time, find talents that match the corresponding crowdsourcing tasks such as capabilities and interests on the Internet to allocate, in order to ensure that the crowdsourcing tasks are completed smoothly and efficiently .

However, in the current crowdsourcing operation scenario, in most cases, when publishing a task, the individual ’s preference for the task is not taken into account. Therefore, after some tasks are distributed, the user does not accept the task in time or does not accept the task. Issues distributed. The user's response to the task is not high, which directly affects the completion degree and delivery efficiency of the task.

Application content

This application provides a crowdsourcing task push method and related device based on user preference, which can improve the delivery efficiency of crowdsourcing tasks.

In the first aspect, this application provides a method for pushing a crowdsourcing task based on user preferences, including:

Determine a first preset number of task response active users for tasks of the target task type;

Obtaining historical push data sets for each task responding to active users separately, the historical push data sets including a second preset amount of historical push data, the historical push data including the historical push data of the target task type that the task responds to the active user being pushed Active user data generated by the task, historical task data corresponding to the historical push task, and response result tag data of whether the historical push task is accepted by the task response active user;

Based on the first preset number of tasks responding to the respective historical push data sets of active users, establishing a historical task acceptance probability logistic regression model of the acceptance probability of the historical task data and the active user data, the historical task acceptance probability Responding to the probability that the active user accepts each historical push task for the task;

Use the acceptance probability logistic regression model to estimate the historical task acceptance probability based on the active user data and the historical task data;

According to the historical task acceptance probability and the response result tag data, estimate the user likelihood degree of the response result corresponding to the historical push task and the response result tag data of the historical push task of the active user by the task response active user;

Determining the first high-preferred target user from the first preset number of task response active users according to the user likelihood corresponding to the first preset number of task response active users, and pushing the target of the target task type The task is pushed to the first high preference target user.

In a second aspect, an embodiment of the present application provides a device for pushing a crowdsourced task based on user preference, including:

An active user determination unit, configured to determine a first preset number of task response active users for tasks of a target task type;

A data set acquiring unit, configured to separately acquire historical push data sets of each task response active user, the historical push data set includes a second preset amount of historical push data, and the historical push data includes the task response active user Active user data generated by a historical push task of the push target task type, historical task data corresponding to the historical push task, and response result tag data of whether the historical push task is accepted by the task in response to an active user;

A model building unit, configured to respond to respective historical push data sets of active users according to the first preset number of tasks, and establish a probability logistic regression model of the acceptance probability of historical tasks regarding the historical task data and the active user data, The acceptance probability of the historical task is the acceptance probability of the task response active user to each historical push task;

An estimation unit, configured to use the acceptance probability logistic regression model to estimate the historical task acceptance probability based on the active user data and the historical task data;

The estimating unit is further configured to estimate the response corresponding to the historical push task and the response result tag data of the historical push task of the active user of the task response based on the historical task acceptance probability and the response result tag data The user likelihood of the result;

A task pushing unit, configured to determine the first target user with the highest preference from the first preset number of task-responsive active users according to the user likelihood corresponding to the first preset number of task-responsive active users, and The target push task of the target task type is pushed to the first high-preference target user.

In a third aspect, an embodiment of the present application provides another apparatus for crowdsourcing task push based on user preference, including a processor, a memory, and a communication interface, where the processor, memory, and communication interface are connected to each other, wherein the communication interface For receiving and sending data, the memory is used to store program code, and the processor is used to call the program code to perform any one of the methods in the first aspect and various possible implementation manners of the first aspect.

According to a fourth aspect, embodiments of the present application provide a computer non-volatile readable storage medium. The computer non-volatile readable storage medium stores a computer program, and the computer program includes program instructions. When being executed by a computer, the computer is caused to execute any one of the above-mentioned first aspect and any possible implementation manner of the first aspect.

In the embodiment of the present application, the crowdsourcing tasks are differentially pushed according to the user's preferences, which improves the user's response degree to the crowdsourcing tasks, thereby improving the delivery efficiency of the crowdsourcing tasks.

BRIEF DESCRIPTION

FIG. 1 is a schematic flowchart of a method for pushing a crowdsourced task based on user preferences according to an embodiment of the present application;

2 is a schematic flowchart of another method for pushing a crowdsourced task based on user preferences provided by an embodiment of the present application;

3 is a schematic structural diagram of a crowd-sourcing task pushing device based on user preferences provided by an embodiment of the present application;

FIG. 4 is a schematic structural diagram of another crowdsourcing task pushing device based on user preferences provided by an embodiment of the present application.

detailed description

The following will describe a method and related apparatus for crowdsourcing task push based on user preference provided by embodiments of the present application in conjunction with FIGS. 1 to 4.

Referring to FIG. 1, FIG. 1 is a schematic flowchart of a method for pushing a crowdsourced task based on user preferences provided by an embodiment of the present application. As shown in the figure, the method includes:

S101: Determine a first preset number of task response active users for tasks of a target task type.

Specifically, the tasks mainly refer to crowdsourcing tasks that appear on the Internet and involve fields such as information retrieval, artificial intelligence, video analysis, knowledge mining, and image quality assessment. There may be multiple task types of the task, for example, semantic judgment type, speech recognition type, information collection type, etc., and the target task type may be one of them. The first preset number can be adjusted according to different tasks of the target task type. It can be understood that the larger the first preset number, that is, the greater the number of active users that the sampled task responds to, through this The user's preferred crowdsourced task pushing method determines the first target user with the highest preference for the task of the target task type is more accurate.

Optionally, the first preset number of active user tasks can be determined by the following steps:

11) Obtain all historical users for the tasks of the target task type respectively. 12) Calculate the target type task acceptance ratio of each historical user separately. The target type task acceptance ratio is the ratio of the number of times each historical user accepts the task of the target task type and the number of times the target task type is pushed. 13) Sort the historical users according to the order of their respective target type task acceptance ratios from high to low, and determine the first preset number of historical users ranked first as the task response active users.

S102: Acquire historical push data sets of each task responding to active users, respectively.

Specifically, for the first preset number of task response active users determined in step S101, a historical push data set of each task response active user is separately collected, and the historical push data set includes a second preset number of historical push data , The historical push data includes active user data generated by the task in response to an active user being pushed by a historical push task of the target task type, historical task data corresponding to the historical push task, and whether the historical push task is used by the task In response to the response result tag data accepted by the active user, that is, the task response active user generates a set of historical push data every time the historical push task is pushed. It can be understood that, the larger the above second preset number, that is, the more historical push data sampled for each task in response to active users, the target task type determined by the crowdsourcing task push method based on user preference The first high preference target user of the task is the more accurate.

Optionally, the historical push information sets of the active users responding to each task are obtained separately, and each historical push information in the historical push information set is quantized to obtain the historical push data sets corresponding to the active users responding to each task. For example, if there is one piece of historical push information in the historical push information set of the acquired task response active user A that the task acceptance time is 17:00, then the historical push information can be quantified as historical push data 17.

S103: According to the first preset number of task response active user's respective historical push data sets, establish a historical task acceptance probability logistic regression model of the acceptance probability of the historical task data and the active user data.

Here, the historical task acceptance probability is the acceptance probability of the task response active user for each historical push task. The acceptance probability logistic regression model is established based on the historical push data set of all the first preset number of task response active users. Using this acceptance probability logistic regression model, any one of the task response active users can be estimated The historical task acceptance probability of the historical push task is described.

Optionally, the historical task data includes at least one task attribute characteristic data of the historical push task, and the active user data includes at least one user attribute characteristic data of the task response active user. If the second preset number is m and m is a positive integer, event A1 is used to indicate that the task responds to the active user u and accepts the i-th historical push task among the m historical push tasks, i is a positive integer and i ∈ [1, m], then the historical task acceptance probability of the task response to the i-th historical push task by the active user u is expressed by formula (1) as:

Among them, the vector x _i is a task feature value vector composed of task attribute feature data corresponding to the ith historical push task; the vector w _u is composed of the user attribute feature data of the task response active user u for the i th historical push task User feature weight vector, the user feature weight of each user attribute feature data in the user feature weight vector, according to all the first preset number of task response active users' respective user attribute feature data corresponding to the task response active user The statistics of the response result label data of the historical push data set are obtained.

S104. Use the acceptance probability logistic regression model to estimate the historical task acceptance probability based on the active user data and the historical task data.

Specifically, using the above-mentioned acceptance probability logistic regression model, according to the first preset number of tasks responding to the active user data and historical task data contained in the respective historical push data sets of the active users, it is estimated that each historical active user has been When pushing the historical push task, the probability of accepting the historical task of the historical push task.

For example, the historical push data set of user A contains two sets of historical push data. These two sets of historical push data correspond to the historical push task of the target task type that was pushed twice by user A. User A was generated when the first historical push task was pushed. Active user data 1, the historical task data of the first historical push task is historical task data 1, user A generates active user data 2 when the second historical push task is pushed, and the historical task data of the second historical push task is historical Task data 2, in this step, using the accepted probability logistic regression model, based on active user data 1 and historical task data 1, it is estimated that when user A was pushed the first historical push task, the historical task of the first historical push task The acceptance probability, based on the active user data 2 and the historical task data 2, estimates the probability of accepting the historical task of the second historical push task when user A was once pushed the second historical push task.

S105: According to the historical task acceptance probability and the response result tag data, estimate the user likelihood of the task response to the active user regarding the historical push task and the response result tag data of the historical push task's response result tag data .

Specifically, according to the task response estimated in step S104, the active user receives the historical task acceptance probability of the historical push task when the historical push task has been pushed by himself, and the response result tag data for the historical push task , Using a likelihood function to estimate the user likelihood of the task response to the active user, the user likelihood is the likelihood of the task response to the active user about their own sampled sample set, the sampled sample set includes the task A historical push task corresponding to all historical task data contained in the historical push data set of the active user and a response result corresponding to all response result tag data contained in the historical push data set of the active user. Since the sampling sample set is a relatively active sampling sample set for the task-responsive active user, that is, the historical push task corresponding to the sampling sample set has been accepted by the corresponding historical active user in a large number, the task can be used to respond to the active The user's degree of likelihood of the user's own sampled set of samples approximates the degree to which the task responds to the active user's preference for the task of the target task type.

Optionally, if the second preset number is m, and m is a positive integer, for the task response to the m historical push tasks pushed by the active user u and the response result corresponding to the response result tag data of the m historical push tasks , Constitutes a sample sample set of the task response active user u, then the likelihood of the task response active user u with respect to the sample sample set is expressed by formula (2) as:

Among them, _Pu (A ₁ | x _uj ) represents the historical task acceptance probability of the jth historical push task among the m historical push tasks that the task responds to the active user u; y _uj is the history of the task response active user u to m The response result label data of the jth historical push task in the push task, j is a positive integer, j ∈ [1, m], y _uj ∈ {0, 1}.

S106: Determine a first high-preferred target user from the first preset number of task response active users according to the user likelihood corresponding to the first preset number of task response active users, and compare The target pushing task is pushed to the target user with the first high preference.

Specifically, since the user likelihood level may indicate the preference level of the task response active user for the task of the target task type, the task with the higher likelihood level estimated by the user in step S105 responds to the target task by the active user The type of task has a higher degree of preference, so a certain number or a certain proportion of tasks with the highest likelihood of the user response to the active user can be determined as the first target user with the highest preference. At this time, it can be pushed to the first high preference target user.

Optionally, the specific step of determining the first high preference target user from the first preset number of task response active users according to the user likelihood corresponding to the first preset number of task response active users It can be as follows:

21) Sort the first preset number of task-responsive active users according to their corresponding user likelihood from high to low. 22). Determine the third preset number of task response active users in the top rank as the first high preference target user, the third preset number is less than or equal to the first preset number, and greater than or equal to The number of the target push tasks.

In an optional embodiment, the method may further include the following steps:

31). Delete the first high preference target user from the first preset number of task response active users, and determine the task response active user after deleting the first high preference target user as a secondary screening active user. 32). When receiving the task rejection instruction of the first high-preferred target user for the target push task, screen the active users from the second level according to the user likelihood corresponding to each of the second-level screening active users The target user with the second highest preference is determined in. 33). Push the target push task corresponding to the task rejection instruction to the second target user with the highest preference.

Through the above step 31), step 32) and step 33), when there is a user who issued a task rejection instruction among the first high preference target users who have been pushed the target push task, the active user can be responded to from the task Among users other than the first high preference target user, a second high preference target user with a higher degree of preference is selected, and a target push task corresponding to the task rejection instruction is pushed to the second high preference target user, both It can be guaranteed that the target push task corresponding to the task rejection instruction matches users with a higher degree of preference, and the first high preference target user can be prevented from being repeatedly pushed by the target push task corresponding to the target task type to reduce the The degree of preference of the first preference target user for the task of the target task type may also increase the exposure of the task of the target task type, so that more second high preference target users than the first high preference target user The task of the target task type is pushed to improve the adhesion between the target user of the second highest preference and the task of the target task type.

In the embodiment of the present application, since the user likelihood degree may represent the preference degree of the task response active user to the task of the target task type, the first high preference target user may be determined according to the user likelihood degree, and then the first high preference The target user pushes a target push task of the target task type. It implements differentiated push crowdsourcing tasks based on user preferences, improves user response to crowdsourcing tasks, and thus improves the delivery efficiency of crowdsourcing tasks.

Referring to FIG. 2, FIG. 2 is a schematic flowchart of another method for pushing a crowdsourced task based on user preference provided by an embodiment of the present application. As shown in the figure, the method includes:

S201: Acquire all historical users of tasks of the target task type respectively.

For example, if the target task type is information collection, all historical users who have been pushed through the information collection task will be acquired.

S202: Calculate the target type task acceptance ratio of each historical user separately, and sort the historical users according to the order of the respective target type task acceptance ratio from high to low, and sort the first preset number of history in the top order The user is determined to be the active user for task response.

Here, the above-mentioned target type task acceptance ratio is the ratio of the number of times each historical user has accepted the above-mentioned target task type task and the number of times the target task type is pushed.

S203: Acquire historical push information sets of active users responding to each task respectively, and quantize each historical push information in the historical push information set to obtain historical push data sets corresponding to active users responding to each task.

Specifically, the historical push information set of each task-responsive active user includes a second preset number of groups of historical push information, and each group of historical push information includes the historical push task of the target task type of the task-responsive active user being pushed The generated active user information, historical task information corresponding to the historical push task, and a response result of whether the historical push task is accepted by the task in response to the active user. The historical task information includes at least one task attribute characteristic information of the historical push task, and the active user information includes at least one user attribute characteristic information of the task response active user. By quantizing the attribute information of each task and the attribute information of each user, the quantitative processing of the historical push information set is completed.

For example, if the target task type is information collection, the first preset number is 50, and the second preset number is 30, the 50 tasks obtained in step S203 respond to user B among active users The historical push information set is shown in Table 1:

Table 1

The collection time, collection channel and information type in Table 1 are the three task characteristic attribute information included in the historical task information, and the age, location and proficiency at the time of being pushed are the three user attribute characteristic information included in the active user information. The above task feature attribute information and each user attribute feature information are quantified, the collection time is quantified as the number corresponding to time, and the collection channel is quantified as: telephone collection, Internet collection and field collection correspond to the numbers 1, 2 and 3, respectively. The information type is quantified as: entertainment type, diet type and medical type correspond to numbers 1, 2 and 3 respectively, quantify age as the number corresponding to age, and quantify the location at the time of being pushed as: home, outdoor and company corresponding numbers 1, 2 and 3, quantify proficiency as: very proficient, more proficient, proficient, less proficient and unskilled, corresponding to the numbers 5, 4, 3, 2 and 1, respectively, and quantify the response result as response result label 1, The response result rejection or omission is quantified as response result label 0, and the historical push data set of user B is obtained, as shown in Table 2:

Table 2

S204: Respond to the user attribute characteristic data and the response result label data contained in the respective historical push data sets of the active users according to all the first preset number of tasks, and count each user attribute characteristic in the user characteristic weight vector composed of the user attribute characteristic data The user characteristic weight of the data.

Specifically, the user feature weight indicates the degree of influence of the corresponding user attribute feature data on the task response to the active user accepting the historical push task. The user feature weight can be calculated through SPSS (Statistical Product and Service Solutions). Service solutions) or SAS (StatisticsAnalysisSystem, statistical analysis system) and other statistical software to calculate.

Optionally, in order to improve the calculation efficiency and reduce the calculation cost, the user attribute characteristic data in the user attribute characteristic data whose absolute value of the corresponding user characteristic weight is less than the first threshold can be discarded to filter out the user attribute characteristic data with a larger user characteristic weight Participate in calculations.

S205: According to the user characteristic weights, establish a logistic regression model of the acceptance probability of historical tasks regarding the historical task data and the active user data.

Based on the example in step S203, if the user feature weights of age, location and proficiency obtained in step S204 are: 0.335, 0.114, and 0.219, define the collection time, collection channel, and collection type using variables t, c and k indicate that age, location when being pushed, and proficiency are respectively expressed by variables a, s, and l, and event A1 is used to indicate that 50 tasks respond to the uth user among active users and receive the ith among 30 historical push tasks For historical push tasks, the u-th user ’s probability of accepting the historical task for the i-th historical push task is expressed by formula (3) as:

Among them, the vector x _i is a task feature value vector composed of the task attribute feature data collection time t _i corresponding to the ith historical push task, the collection channel c _i , and the information type k _i , and the vector x _i = (t _i , c _i , K _i); vector w _u to u th user's properties for the i-th history push task feature data Age a _u, when pushed location S _u, a user feature weight vector, vector w proficiency L _u configuration _u = (0.335 * a _u , 0.114 * s _u , 0.219 * l _u )

S206. Use the acceptance probability logistic regression model to estimate the historical task acceptance probability based on the active user data and the historical task data.

Here, the historical task acceptance probability is the probability that the task response active user receives the historical push task corresponding to the historical task data when the historical push task corresponding to the historical task data has been pushed.

Based on the example in step S203, if the user characteristic weights of age, location and proficiency calculated by the statistical software SPSS in step S204 are: 0.335, 0.114 and 0.219, according to the historical push data set corresponding to Table 2 , The task feature value vector of the historical push task corresponding to label 3 is x ₃ = (12,3,3), and the user feature weight vector of user B for the historical push task corresponding to label 3 is w _B = (0.335 * 20,0.114 * 3,0.219 * 4), substituting x ₃ and w _B into the formula (3) in step S205 to calculate that user B has been pushed to the historical push task corresponding to the number 3, and the number is 3 The historical task acceptance probability P _B (A ₁ | x ₃ ) of the corresponding historical push task. According to the above method, when user B is pushed the historical push tasks corresponding to 1, 2, 3, ..., 30, respectively, the historical push tasks corresponding to 1, 2, 3, ..., 30 can be estimated. Historical task acceptance probabilities P _B (A ₁ | x ₁ ), P _B (A ₁ | x ₂ ), P _B (A ₁ | x ₃ ), ..., P _B (A ₁ | x ₃₀ ). Furthermore, it can be estimated that each task responds to the historical task of the corresponding historical push task when the active user has been pushed the corresponding historical push task.

S207, according to the historical task acceptance probability and the response result tag data, estimate the user likelihood degree of the response result corresponding to the historical push task and the response result tag data of the historical push task by the active user of the task response to the historical push task .

Based on the example in step S203, if it is obtained in step S206 that user B is pushed a historical push task corresponding to labels 1, 2, 3, ..., 30, respectively, corresponding to the labels 1, 2, 3, ..., 30 The historical task acceptance probabilities of historical push tasks P _B (A ₁ | x ₁ ), P _B (A ₁ | x ₂ ), P _B (A ₁ | x ₃ ), ..., P _B (A ₁ | x ₃₀ ), User B's sampling sample set is [(historical push task corresponding to label 1, accepted), (historical push task corresponding to label 2, rejected), (historical push task corresponding to label 3, accepted), (labeled as 4 The corresponding historical push task, ignore), ..., (the historical push task corresponding to 30, accept)], the user likelihood of user B's own sample set can be calculated using formula (4):

P (y _B | w _B ) = P _B (A ₁ | x ₁ ) * [1-P _B (A ₁ | x ₂ )] * [1-P _B (A ₁ | x ₃ )] *… * P _B (A ₁ | x ₃₀ ) ... Formula (4) calculates the likelihood of 50 task response samples sampled by the active user in turn according to the above method.

S208: Sort the first preset number of task-responsive active users according to their respective user likelihood levels from high to low, and determine the third highest preset number of task-responsive active users as the first High preference for target users.

Here, the third preset number is the number of the first high preference target users, and the value of the third preset number may be a preset fixed value, or may be the set first preset A fixed percentage of the value of the quantity, the fixed value and the fixed percentage can be set differently according to different attribute information such as the emergency of the target push task, but the setting of the fixed value or the fixed percentage must ensure that The third preset number is less than or equal to the first preset number, and greater than or equal to the number of target push tasks.

For example, if the 50 tasks responding to active users in step S207 are sorted according to their respective user likelihood from high to low, the sequence is obtained as shown in Table 3. If the third preset number is 3, the user 20, User 8 and user 44 are determined to be the first target users with the highest preference. Table 3 is as follows:

Sort label	user	User likelihood
1	User 20	0.0835
2	User 8	0.0762
3	User 44	0.0754
...	...	...
50	User 39	0.0014

table 3

S209: Determine the high-preference acceptance time data of the first high-preference target user according to the historical push data set of the first high-preference target user, and report to the first at the time corresponding to the high-preference acceptance time data A high preference target user pushes a target push task of the target task type.

Specifically, the time corresponding to the high preference acceptance time data is the time when the first high preference target user is most likely to accept the target push task in 24 hours of the day, and the first high preference target user accepts the Pushing at the time when the target push task is most likely can further increase the possibility that the first high-preference target user accepts the target push task, thereby improving the delivery efficiency of the target push task.

Here, it is possible to determine one piece of high preference acceptance time data common to the first high preference target user, and to push the target push task to the first high preference target user in a unified manner at the moment corresponding to the high preference acceptance time, or Determine the respective high preference acceptance time data of each first high preference target user, and push the target push task to the corresponding first high preference target user at the time corresponding to the respective high preference acceptance time data of each first high preference target user Furthermore, there are multiple methods for determining the high-preference acceptance time data of the first high-preference target user.

In an optional implementation manner, statistical calculations can be performed on the historical push data sets of all the first high-preferred target users, and the results are obtained at 0 o'clock-1 o'clock, 1 o'clock-2 o'clock, 2 o'clock-3 o'clock, ... 2. The reception peak time period with the largest number of accepted historical push tasks for pushes between 23:00 and 22:00 can be determined as the high preference reception time data.

Each historical push data set of the first high preference target user may also be separately calculated to obtain respective peak reception periods of each first high preference target user, and the respective peak reception times of the first high preference target users The time period pushes the target push task to the corresponding first high preference target user.

In another optional embodiment, the fourth preset number of historical task acceptance probabilities that are the highest among the historical task acceptance probabilities of all the first high preference target users are acquired, and the history of the highest fourth preset number of The push time data of the historical push task corresponding to the task acceptance probability is determined to be the high preference acceptance time data, and the target push task of the target task type is assigned to a fourth preset number of push batches, respectively in the fourth A target push task corresponding to a push batch is pushed to the first high-preference target user at a time corresponding to a preset number of high-preference acceptance time data. The push time data may exist in active user data included in the historical push data set, or may exist in historical task data included in the historical push data set.

For example, if the fourth preset number is 3, the highest three historical task acceptance probabilities among the historical task acceptance probabilities of all the first high-priority target users are the third target that User A has pushed to himself. The historical task acceptance probability P _A (A ₁ | x ₃ ) of the historical push task of the task type, the user M's historical task acceptance probability P _M (A ₁ | x ₁₂ ). User G has a history task acceptance probability P _G (A ₁ | x ₂₉ ) for the history push task of the 29th target task type that he has been pushed. Further, it can be pushed according to the history of user A The data set determines that the time when user A was pushed the third historical push task is 17:00, and the time according to user M's historical push data set determines that user M was once pushed the twelfth historical push task as 11. Point, according to the historical push data set of user G, it is determined that the time when user G was pushed the 29th historical push task is 15 o'clock. If there are 15 target push tasks in total, randomize these 15 target push tasks 3 push batches, for example, you can randomly assign 5 target push tasks to each push batch, and then push the 3 pushes to the first high preference target user at 11:00, 15:00 and 17:00 respectively Batch target push tasks.

It is also possible to obtain the respective highest historical task acceptance probabilities of the first high preference target users respectively, and further determine the push moments of the historical push tasks corresponding to the respective highest historical task acceptance probabilities of the first high preference target users, respectively Said the pushing moment of the historical push task corresponding to the highest probability of accepting the historical task of the first high preference target user, and pushing the target push task to the first high preference target user.

Optionally, if a task rejection instruction of the first high preference target user for the target push task is received, determine a second high preference target user for the target push task corresponding to the task rejection instruction, and further determine the The second high preference target user's high preference acceptance time data, and at the moment corresponding to the second high preference target user's high preference acceptance time data, push the target push task corresponding to the task rejection instruction to the second highest Preference target users.

Based on the example in step S208, if the target push tasks of 6 target task types are pushed to user 20, user 8, user 44, user 6, user 29, and user 12, the task rejection instructions of user 44 and user 12 are received , Then user 3 and user 1 can be determined as the second-highest-priority target users, and then the highest historical task acceptance probability of user 3 and user 1 are obtained respectively as the 18th target that user 3 has pushed to itself The historical task acceptance probability of the historical push task of the task type and the historical task acceptance probability of the historical push task of the fifth target task type that user 1 has been pushed by himself, and the 18th historical push of user 3 who was pushed The task time is 3 o'clock, and the time when user 1 was once pushed the fifth historical push task is 10 o'clock, then the historical push task corresponding to the task rejection instruction sent by user 44 and user 12 respectively is at 3 o'clock and 10 Point push to user 3 and user 1.

In the embodiments of the present application, not only is the differentiated push of crowdsourcing tasks based on the user's preference implemented, but also when the user accepts the task, the task is pushed to the task, which further improves the user's response to the crowdsourcing task and improves Delivery efficiency of crowdsourcing tasks.

Referring to FIG. 3, FIG. 3 is a schematic structural diagram of a device for crowdsourcing task push based on user preference provided by an embodiment of the present application. As shown in the figure, the device 30 for crowdsourcing task push based on user preference may include:

The active user determination unit 301 is configured to determine a first preset number of task response active users for tasks of a target task type.

Specifically, there may be multiple task types of the task, for example, semantic judgment type, speech recognition type, information collection type, etc., and the target task type may be one of them. The first preset number can be adjusted according to different tasks of the target task type. It can be understood that the larger the first preset number is, the greater the number of active users that the sampled task responds to. The more accurate the target user with the first highest preference for the task of the target task type determined by the preferred crowdsourcing task pushing device.

Optionally, the active user determining unit 301 may be specifically configured to: separately obtain all historical users of the task for the target task type; calculate the target type task acceptance ratio of each historical user separately, and the target type task acceptance ratio The ratio of the number of times each historical user accepts the task of the above target task type and the number of times the target task type is pushed; sort the historical users according to the order of acceptance of the respective target type task from high to low, The first preset number of historical users ranked top is determined to be the task response active user.

The data set acquiring unit 302 is configured to separately acquire historical push data sets of each task responding to active users, the historical push data set includes a second preset amount of historical push data, and the historical push data includes the task response to the active user being pushed to the target task Active user data generated by a type of historical push task, historical task data corresponding to the historical push task, and response result tag data whether the historical push task is accepted by the task in response to an active user.

Specifically, for the first number of task response active users determined by the active user determination unit 301, a historical push data set of each task response active user is separately collected, and the historical push data set includes a second preset number of historical push data , The historical push data includes active user data generated by the task in response to an active user being pushed by a historical push task of the target task type, historical task data corresponding to the historical push task, and whether the historical push task is used by the task In response to the response result tag data accepted by the active user, that is, the task response active user generates a set of historical push data every time the historical push task is pushed. It can be understood that the larger the above-mentioned second preset number, that is, the more historical push data sampled for each task in response to the active user, the crowd-sourcing task push device based on the user preference determines the target task type for the target task type. The target user with the highest preference for the task is more accurate.

Optionally, the data set acquiring unit 302 may specifically acquire historical push information sets of active users responding to each task, and quantize each historical push information in the historical push information set to obtain correspondences between active users responding to each task Historical push data set.

The model building unit 303 is configured to respond to the respective historical push data sets of active users according to the first preset number of tasks, and establish a logistic regression model of the probability of acceptance of historical tasks with respect to the historical task data and the data of active users .

Here, the historical task acceptance probability is the acceptance probability of the task response active user for each historical push task. The acceptance probability logistic regression model is established by the model building unit 303 based on all the first preset number of task response historical push data sets of active users, and the acceptance probability logistic regression model can be used to estimate any of the task responses The probability of an active user accepting a historical task of any of the historical push tasks.

Optionally, the historical task data includes at least one task attribute characteristic data of the historical push task, and the active user data includes at least one user attribute characteristic data of the task response active user. If the second preset number is m and m is a positive integer, event A1 is used to indicate that the task responds to the active user u and accepts the i-th historical push task among the m historical push tasks, i is a positive integer and i∈ [1, m], then the model building unit 303 establishes a logistic regression model of acceptance probability, and the historical task acceptance probability of the task response active user u to the ith historical push task is expressed by formula (5) as:

Among them, the vector x _i is a task feature value vector composed of task attribute feature data corresponding to the ith historical push task; the vector w _u is composed of the user attribute feature data of the task response active user u for the i th historical push task User feature weight vector, user feature weight of each user attribute feature data in the user feature weight vector, the model building unit 303 responds to the respective user attribute feature data and all of the active users according to all the first preset number of tasks The said task is statistically obtained by responding to the response result label data of the historical push data set corresponding to the active user.

The estimating unit 304 is configured to use the acceptance probability logistic regression model to estimate the historical task acceptance probability based on the active user data and the historical task data.

Specifically, the estimation unit 304 uses the acceptance probability logistic regression model to respond to the active user data and historical task data contained in the respective historical push data sets of the active users according to the first preset number of tasks, and estimates that each historical active user has been When pushing the historical push task, the probability of accepting the historical task of the historical push task.

The estimating unit 304 is further configured to estimate, according to the historical task acceptance probability and the response result tag data, the task response active user corresponding to the historical push task and the response result tag data of the historical push task The likelihood of the user responding to the results.

Specifically, the estimation unit 304 responds to the historical task of the historical push task when the active user responds to the historical push task that has been pushed according to the estimated task, and the response result for the historical push task Tag data, using a likelihood function to estimate the user likelihood of the task-responsive active user, the user likelihood is the likelihood of the task-responsive active user with respect to their sampled sample set, the sampled sample set includes all The historical response task corresponding to all historical task data contained in the historical push data set of the task response active user and the response result corresponding to all response result tag data contained in the historical response data set of the active user response. Since the sampling sample set is a relatively active sampling sample set for the task-responsive active user, that is, the historical push task corresponding to the sampling sample set has been accepted by the corresponding historical active user in a large number, the task can be used to respond to active The user's degree of likelihood of the user's own sampled set of samples approximates the degree to which the task responds to the active user's preference for the task of the target task type.

Optionally, if the second preset number is m, and m is a positive integer, for the task response to the m historical push tasks pushed by the active user u and the response result corresponding to the response result tag data of the m historical push tasks , Constitutes a sample sample set of the task response active user u, then the estimation unit 304 expresses the likelihood of the task response active user u with respect to the sample sample set as formula (6) as:

Among them, _Pu (A ₁ | x _uj ) represents the historical task acceptance probability of the jth historical push task among the m historical push tasks that the task responds to the active user u; y _uj is the history of the task response active user u to m The response result label data of the jth historical push task in the push task, j is a positive integer, j ∈ [1, m], y _uj ∈ {0, 1}.

A task pushing unit 305, configured to determine the first high-preferred target user from the first preset number of task-responsive active users according to the user likelihood corresponding to the first preset number of task-responsive active users, and Push the target push task of the target task type to the first high-preferred target user.

Specifically, since the user likelihood level may indicate the preference level of the task-responsive active user for the task of the target task type, the task-responsive active user with a higher task likelihood level estimated by the estimation unit 304 is The task of the target task type has a higher degree of preference. Therefore, a certain number or a certain percentage of the task response active users with the highest likelihood of the user may be determined as the first target user with the highest preference. When there is a target push task of the target task type When it is to be pushed, it may be pushed to the first target user with the highest preference.

Optionally, the task pushing unit 305 may be specifically configured to: sort the first preset number of tasks in response to active users according to their corresponding user likelihood levels from high to low; and sort the top third A preset number of tasks is determined to be the first high preference target user in response to the active user, and the third preset number is less than or equal to the first preset number and greater than or equal to the number of target push tasks.

Further optionally, the crowdsourcing task pushing device 30 based on user preferences may further include:

The secondary filtering unit 306 is configured to delete the first high preference target user from the first preset number of task response active users, and determine the task response active user after deleting the first high preference target user Screen active users for the second level.

The secondary screening unit 306 is further configured to, when receiving the task rejection instruction of the first high-preferred target user for the target push task, according to the user likelihood corresponding to the secondary screening of the active users, from all Among the second-level screening active users, the second highest preference target user is determined.

The task pushing unit 305 is further configured to push the target pushing task corresponding to the task rejection instruction to the second high-preferred target user.

When there is a user who issued a task rejection instruction among the first high preference target users who have been pushed the target push task, the secondary filtering unit 306 may respond to the first high preference among active users from the task A second high preference target user with a higher degree of preference is selected from users other than the target user, and the task pushing unit 305 pushes the target push task corresponding to the task rejection instruction to the second high preference target user, either It is guaranteed that the target push task corresponding to the task rejection instruction matches users with a higher degree of preference, and the first high preference target user can be prevented from being repeatedly pushed to the target push task corresponding to the target task type and reduce the first A preference target user's preference for tasks of the target task type can also increase the exposure of the task of the target task type, so that more second high preference target users than the first high preference target user are Pushing the task of the target task type to improve the adhesion of the target user with the second highest preference to the task of the target task type.

Further optionally, the task pushing unit 305 may be specifically configured to: according to the historical push data set of the first high preference target user, determine the high preference acceptance time data of the first high preference target user; The time corresponding to the high preference acceptance time data pushes the target push task of the target task type to the first high preference target user.

Specifically, the time corresponding to the high preference acceptance time data is the time when the first high preference target user has the highest probability of accepting the target push task in 24 hours of the day, and the task pushing unit 305 is at the first high preference The target user accepts the push at the time when the probability of the target push task is maximized, which can further increase the possibility that the first high-preference target user accepts the target push task, thereby improving the delivery efficiency of the target push task. Here, the task pushing unit 305 may determine one piece of high preference acceptance time data common to the first high preference target user, and uniformly push the first high preference target user at the time corresponding to the high preference acceptance time The target push task can also determine the respective high preference acceptance time data of each first high preference target user, and send the corresponding first high preference target user at the time corresponding to the respective high preference acceptance time data of each first high preference target user Push the target push task.

In the embodiment of the present application, since the user likelihood degree may indicate the preference degree of the task response active user to the task of the target task type, the first high preference target user may be determined according to the user likelihood degree, and then the first high preference target user The target push task of the push target task type. It implements differentiated push crowdsourcing tasks based on user preferences, improves user response to crowdsourcing tasks, and thus improves the delivery efficiency of crowdsourcing tasks.

Referring to FIG. 4, FIG. 4 is a schematic structural diagram of another device for crowdsourcing task push based on user preference provided by an embodiment of the present application. As shown in the figure, the device 40 for crowdsourcing task push based on user preference includes a processor 401 、 Memory 402 and communication interface 403. The processor 401 is connected to the memory 402 and the communication interface 403. The memory 402 is used to store a computer program, and the computer program includes program instructions, and the processor 401 is used to execute the program instructions stored in the memory 402. Wherein, the processor 401 is configured to call the program instruction to execute:

Determine a first preset number of task response active users for tasks of the target task type;

Obtaining historical push data sets for each task responding to active users separately, the historical push data sets including a second preset amount of historical push data, the historical push data including the historical push data of the target task type that the task responds to the active user being pushed Active user data generated by the task, historical task data corresponding to the historical push task, and response result tag data of whether the historical push task is accepted by the task response active user;

Based on the first preset number of tasks responding to the respective historical push data sets of active users, establishing a historical task acceptance probability logistic regression model of the acceptance probability of the historical task data and the active user data, the historical task acceptance probability Responding to the probability that the active user accepts each historical push task for the task;

Use the acceptance probability logistic regression model to estimate the historical task acceptance probability based on the active user data and the historical task data;

According to the historical task acceptance probability and the response result tag data, estimate the user likelihood degree of the response result corresponding to the historical push task and the response result tag data of the historical push task of the active user by the task response active user;

Determining the first high-preferred target user from the first preset number of task response active users according to the user likelihood corresponding to the first preset number of task response active users, and pushing the target of the target task type The task is pushed to the first high preference target user.

It should be understood that in the embodiment of the present application, the processor 401 may be a central processing unit (Central Processing Unit, CPU), and the processor may also be other general-purpose processors, digital signal processors (Digital Signal Processors, DSP), dedicated Integrated circuit (Application Specific Integrated Circuit, ASIC), ready-made programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The memory 402 may include read-only memory and random access memory, and provide instructions and data to the processor 401. A portion of the memory 402 may also include non-volatile random access memory. For example, the memory 402 may also store device type information.

It should be noted that the implementation of each operation may also correspond to the corresponding description of the method embodiments shown in FIGS. 1-2; the processor 401 may also be used to perform other operations in the above method embodiments.

An embodiment of the present application further provides a computer non-volatile readable storage medium, the computer non-volatile readable storage medium stores a computer program, the computer program includes program instructions, and the program instructions are executed by a computer When the computer executes the method as described in the foregoing embodiment, the computer may be a part of the above-mentioned crowd-sourced task pushing device based on user preference.

A person of ordinary skill in the art may understand that all or part of the processes in the method of the foregoing embodiments may be completed by instructing relevant hardware through a computer program, and the program may be stored in a computer-readable storage medium. During execution, the process of the above method embodiments may be included. Wherein, the storage medium may be a magnetic disk, an optical disk, a read-only memory (Read-Only Memory, ROM) or a random access memory (Random Access Memory, RAM), etc.

The above is only the specific implementation of this application, but the scope of protection of this application is not limited to this, any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in this application. It should be covered by the scope of protection of this application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims (20)

1. A method for pushing crowdsourcing tasks based on user preferences is characterized by including:

   Determine a first preset number of task response active users for tasks of the target task type;

   Obtaining historical push data sets for each task responding to active users separately, the historical push data sets including a second preset amount of historical push data, the historical push data including the historical push data of the target task type that the task responds to the active user being pushed Active user data generated by the task, historical task data corresponding to the historical push task, and response result tag data of whether the historical push task is accepted by the task response active user;

   Based on the first preset number of tasks responding to the respective historical push data sets of active users, establishing a historical task acceptance probability logistic regression model of the acceptance probability of the historical task data and the active user data, the historical task acceptance probability Responding to the probability that the active user accepts each historical push task for the task;

   Use the acceptance probability logistic regression model to estimate the historical task acceptance probability based on the active user data and the historical task data;

   According to the historical task acceptance probability and the response result tag data, estimate the user likelihood degree of the response result corresponding to the historical push task and the response result tag data of the historical push task of the active user by the task response active user;

   Determining the first high-preferred target user from the first preset number of task response active users according to the user likelihood corresponding to the first preset number of task response active users, and pushing the target of the target task type The task is pushed to the first high preference target user.
2. The method according to claim 1, wherein the historical task data includes at least one task attribute characteristic data of the historical push task, and the active user data includes at least one user attribute characteristic of the task response active user data;

   The responding to the respective historical push data sets of the active users according to the first preset number of tasks, and establishing a probability of receiving a historical task logistic regression model of the probability of acceptance of the historical task data and the active user data includes:

   If the second preset number is m and m is a positive integer, event A1 is used to indicate that the task responds to the active user u and accepts the i-th historical push task among the m historical push tasks, i is a positive integer and i ∈ [1, m], the historical task acceptance probability of the task response to the i-th historical push task by the active user u is:

   

   Among them, the vector x _i is a task feature value vector composed of task attribute feature data corresponding to the ith historical push task; the vector w _u is composed of the user attribute feature data of the task response active user u for the i th historical push task User feature weight vector, the user feature weight of each user attribute feature data in the user feature weight vector, according to all the first preset number of task response active users' respective user attribute feature data corresponding to the task response active user The statistics of the response result label data of the historical push data set are obtained.
3. The method according to claim 1 or 2, characterized in that, based on the historical task acceptance probability and the response result tag data, the task response is estimated to respond to active users regarding the historical push task and the historical push The user likelihood of the response result corresponding to the response result tag data of the task includes:

   If the second preset number is m and m is a positive integer, the response results corresponding to the m historical push tasks pushed by the active user u and the response result label data of the m historical push tasks for the task constitute the If the task responds to the sampled sample set of the active user u, then the likelihood that the task responds to the active user u about the sampled sample set is:

   

   Among them, _Pu (A ₁ | x _uj ) represents the historical task acceptance probability of the jth historical push task among the m historical push tasks that the task responds to the active user u; y _uj is the history of the task response active user u to m The response result label data of the jth historical push task in the push task, j is a positive integer, j ∈ [1, m], y _uj ∈ {0, 1}.
4. The method according to any one of claims 1 to 3, wherein the response to the user likelihood corresponding to each active user according to the first preset number of tasks is from the first preset number of tasks Among the active users, the first target user with the highest preference includes:

   Sort the first preset number of tasks in response to active users according to their corresponding user likelihood from high to low;

   A third preset number of task responding active users ranked at the top is determined as the first high preference target user, the third preset number is less than or equal to the first preset number and greater than or equal to the target The number of push tasks.
5. The method according to any one of claims 1 to 4, wherein the method further comprises:

   Deleting the first high preference target user from the first preset number of task response active users, and determining the task response active user after deleting the first high preference target user as a secondary screening active user;

   When receiving the task rejection instruction of the first high preference target user with respect to the target push task, determine the first Second high preference target users;

   Pushing the target push task corresponding to the task rejection instruction to the second high-preferred target user.
6. The method according to any one of claims 1-5, wherein the pushing the target push task of the target task type to the first high-preferred target user includes:

   Determining the high preference acceptance time data of the first high preference target user according to the historical push data set of the first high preference target user;

   The target push task of the target task type is pushed to the first high preference target user at the moment corresponding to the high preference acceptance time data.
7. The method according to any one of claims 1-6, wherein the determining a first preset number of task response active users for the task of the target task type includes:

   Acquiring all historical users corresponding to the task of the target task type;

   Calculate the target type task acceptance ratio of each of the historical users separately, the target type task acceptance ratio is the number of times each historical user accepts the task of the target task type and the number of times the target task type is pushed ratio;

   The historical users are sorted in order from the highest to low task acceptance ratios of their respective target types, and the first preset number of historical users that are ranked first are determined as the task response active users.
8. The method according to any one of claims 1-7, wherein the separately acquiring historical push data sets of each task in response to active users includes:

   Obtain the historical push information set of each task responding to active users separately;

   Quantify each historical push information in the historical push information set to obtain a historical push data set corresponding to the active user responding to each task.
9. The method according to any one of claims 2, wherein the method further comprises:

   Based on all the first preset number of task response active user's respective user attribute characteristic data, and the task response active user's corresponding response result label data of the historical push data set, statistically obtain the user characteristic weight of each user attribute characteristic data;

   In the user attribute characteristic data of the task response active user u for the i-th historical push task, the user attribute characteristic data whose user characteristic weight is not less than the first threshold constitutes a vector w _u .
10. A crowdsourcing task pushing device based on user preference is characterized in that it includes:

    An active user determination unit, configured to determine a first preset number of task response active users for tasks of a target task type;

    A data set acquiring unit, configured to separately acquire historical push data sets of each task response active user, the historical push data set includes a second preset amount of historical push data, and the historical push data includes the task response active user Active user data generated by a historical push task of the push target task type, historical task data corresponding to the historical push task, and response result tag data of whether the historical push task is accepted by the task in response to an active user;

    A model building unit, configured to respond to respective historical push data sets of active users according to the first preset number of tasks, and establish a probability logistic regression model of the acceptance probability of historical tasks regarding the historical task data and the active user data, The acceptance probability of the historical task is the acceptance probability of the task response active user to each historical push task;

    An estimation unit, configured to use the acceptance probability logistic regression model to estimate the historical task acceptance probability based on the active user data and the historical task data;

    The estimating unit is further configured to estimate the response corresponding to the historical push task and the response result tag data of the historical push task of the active user of the task response based on the historical task acceptance probability and the response result tag data The user likelihood of the result;

    A task pushing unit, configured to determine the first target user with the highest preference from the first preset number of task-responsive active users according to the user likelihood corresponding to the first preset number of task-responsive active users, and The target push task of the target task type is pushed to the first high-preference target user.
11. The apparatus according to claim 10, wherein the historical task data includes at least one task attribute characteristic data of the historical push task, and the active user data includes at least one user attribute characteristic of the task response active user data;

    If the second preset number is m and m is a positive integer, event A1 is used to indicate that the task responds to the active user u and accepts the i-th historical push task among the m historical push tasks, i is a positive integer and i ∈ [1, m], then the model building unit is specifically configured to determine the historical task acceptance probability of the task in response to the active user u for the ith historical push task as:

    

    Among them, the vector x _i is a task feature value vector composed of task attribute feature data corresponding to the ith historical push task; the vector w _u is composed of the user attribute feature data of the task response active user u for the i th historical push task User feature weight vector, the user feature weight of each user attribute feature data in the user feature weight vector, according to all the first preset number of task response active users' respective user attribute feature data corresponding to the task response active user The statistics of the response result label data of the historical push data set are obtained.
12. The device according to claim 10 or 11, wherein, if the second preset number is m, and m is a positive integer, m tasks in response to tasks are pushed by the m user history pushed by the active user u and the m history The response result corresponding to the response result tag data of the push task constitutes the sampling sample set of the task response active user u, and the estimation unit is specifically used to determine the likelihood of the task response active user u with respect to the sampling sample set for:

    

    Among them, _Pu (A ₁ | x _uj ) represents the historical task acceptance probability of the jth historical push task among the m historical push tasks that the task responds to the active user u; y _uj is the history of the task response active user u to m The response result label data of the jth historical push task in the push task, j is a positive integer, j ∈ [1, m], y _uj ∈ {0, 1}.
13. The device according to any one of claims 10-12, wherein the task pushing unit is specifically used to:

    Sort the first preset number of tasks in response to active users according to their corresponding user likelihood from high to low;

    A third preset number of task responding active users ranked at the top is determined as the first high preference target user, the third preset number is less than or equal to the first preset number and greater than or equal to the target The number of push tasks.
14. The device according to any one of claims 10-13, wherein the device further comprises:

    A secondary filtering unit, configured to delete the first high preference target user from the first preset number of task response active users, and determine the task response active user after deleting the first high preference target user as Secondary screening of active users;

    The secondary screening unit is further configured to, when receiving the task rejection instruction of the first high-preferred target user for the target push task, according to the user likelihood corresponding to the secondary screening active users, from the Identify the second highest preference target users among the second-level screening active users;

    The task pushing unit is further configured to push the target pushing task corresponding to the task rejection instruction to the second high-preferred target user.
15. The method according to any one of claims 10 to 14, wherein the task pushing unit is specifically used to:

    Determining the high preference acceptance time data of the first high preference target user according to the historical push data set of the first high preference target user;

    The target push task of the target task type is pushed to the first high preference target user at the moment corresponding to the high preference acceptance time data.
16. A crowdsourcing task pushing device based on user preference is characterized by comprising a processor, a memory and a communication interface, the processor, the memory and the communication interface are connected to each other, wherein the communication interface is used to receive and send data, The memory is used to store program code, and the processor is used to call the program code and execute:

    Determine a first preset number of task response active users for tasks of the target task type;

    Obtaining historical push data sets for each task responding to active users separately, the historical push data sets including a second preset amount of historical push data, the historical push data including the historical push data of the target task type that the task responds to the active user being pushed Active user data generated by the task, historical task data corresponding to the historical push task, and response result tag data of whether the historical push task is accepted by the task response active user;

    Based on the first preset number of tasks responding to the respective historical push data sets of active users, establishing a historical task acceptance probability logistic regression model of the acceptance probability of the historical task data and the active user data, the historical task acceptance probability Responding to the probability that the active user accepts each historical push task for the task;

    Use the acceptance probability logistic regression model to estimate the historical task acceptance probability based on the active user data and the historical task data;

    According to the historical task acceptance probability and the response result tag data, estimate the user likelihood degree of the response result corresponding to the historical push task and the response result tag data of the historical push task of the active user by the task response active user;

    Determining the first high-preferred target user from the first preset number of task response active users according to the user likelihood corresponding to the first preset number of task response active users, and pushing the target of the target task type The task is pushed to the first high preference target user.
17. The apparatus according to claim 16, wherein the historical task data includes at least one task attribute characteristic data of the historical push task, and the active user data includes at least one user attribute characteristic of the task response active user data;

    The processor is specifically used for:

    In the second preset number is m, m is a positive integer, and event A1 is used to indicate that the task responds to the active user u to accept the i-th historical push task among the m historical push tasks, i is a positive integer, i∈ [1, In the case of m], it is determined that the task responds to the active user u and the historical task acceptance probability of the i-th historical push task is:

    

    Among them, the vector x _i is a task feature value vector composed of task attribute feature data corresponding to the ith historical push task; the vector w _u is composed of the user attribute feature data of the task response active user u for the i th historical push task User feature weight vector, the user feature weight of each user attribute feature data in the user feature weight vector, according to all the first preset number of task response active users' respective user attribute feature data corresponding to the task response active user The statistics of the response result label data of the historical push data set are obtained.
18. The apparatus according to claim 16 or 17, wherein the processor is specifically configured to:

    When the second preset number is m, and m is a positive integer, the response results corresponding to the m historical push tasks pushed by the active user u and the response result tag data of the m historical push tasks for the task constitute the In the case that the task responds to the sampled sample set of the active user u, the likelihood degree that the task responds to the active user u about the sampled sample set is:

    

    Among them, _Pu (A ₁ | x _uj ) represents the historical task acceptance probability of the jth historical push task among the m historical push tasks that the task responds to the active user u; y _uj is the history of the task response active user u to m The response result label data of the jth historical push task in the push task, j is a positive integer, j ∈ [1, m], y _uj ∈ {0, 1}.
19. The device according to any one of claims 16-18, wherein the processor is specifically configured to:

    Sort the first preset number of tasks in response to active users according to their corresponding user likelihood from high to low;

    A third preset number of task responding active users ranked at the top is determined as the first high preference target user, the third preset number is less than or equal to the first preset number and greater than or equal to the target The number of push tasks.
20. A computer non-volatile readable storage medium, characterized in that the computer non-volatile readable storage medium stores a computer program, and the computer program includes program instructions, which are executed by a processor Causing the processor to perform the method according to any one of claims 1-9.

Images