Optimal worker selection method and system based on reverse backpack in abnormal data crowdsourcing detection

Abstract

The present invention belongs to the technical field of crowdsourcing network, and particularly relates to a method and system for selecting optimal workers based on reverse backpack in abnormal data crowdsourcing detection. The method and system determine the set of bidding workers according to the abnormal data detection task and the crowdsourcing worker information in the crowdsourcing network, obtain multiple worker combinations according to the abnormal data detection task requirements and add them to the candidate crowdsourcing worker set; obtain the reliability of the prediction of the abnormal data detection result by weighted calculation of the crowdsourcing worker trust value, use the total bidding price of the crowdsourcing worker as a constraint condition, the candidate crowdsourcing worker set as a backpack item, the crowdsourcing worker trust value as the backpack item weight, and the expected reliability threshold as the backpack capacity, and model the worker selection problem in abnormal data detection as a reverse backpack model; and select the optimal worker combination by solving the reverse backpack model. The present invention comprehensively considers the expected reliability of the detection task and the bidding price to select the corresponding optimal worker combination, saving labor and time costs.

Description

Technical Field

[0001] The present invention belongs to the technical field of crowdsourcing networks, and in particular relates to a method and system for selecting optimal workers based on a reverse backpack in crowdsourcing detection of abnormal data. Background Art

[0002] Abnormal data detection is widely used in finance, aerospace, medical and other fields. It can ensure the authenticity and availability of data. Although there are many research results on big data and there are many abnormal data detection methods based on machine learning or mathematical statistics, many fields such as social computing and social public security have put forward high requirements for big data computing performance. When dealing with data anomaly detection problems with large data scale, the computing power of a single node is not enough to meet the timeliness of user needs, resulting in low detection efficiency and the reliability of detection cannot be guaranteed. Therefore, how to efficiently and reliably process abnormal data in a large amount of data and improve the availability of data has become a key link in information construction. Crowdsourcing network is a task allocation method in which an individual or organization outsources work tasks to the public network. Users of crowdsourcing network are divided into two categories: users who publish crowdsourcing tasks are called crowdsourcing task publishers, and users who execute crowdsourcing tasks are called crowdsourcing workers. After completing crowdsourcing tasks, crowdsourcing workers can receive corresponding rewards. Crowdsourcing network helps task publishers in crowdsourcing network solve a large number of complex problems by gathering large-scale intelligence. Common crowdsourcing tasks include information collection, image labeling, large-scale computing, etc. Combining the crowdsourcing method with abnormal data detection, dividing large-scale complex tasks and selecting appropriate work nodes for allocation can effectively improve the efficiency of abnormal data detection. However, the openness of the crowdsourcing network makes the reliability of workers in the crowdsourcing network uneven, resulting in a large number of wrong answers in the crowdsourcing task results, reducing the reliability of the crowdsourcing task results. At the same time, the mismatch between the bidding price of abnormal data detection and the reliability of the results leads to higher crowdsourcing task costs and lower results reliability. Therefore, how to reasonably select crowdsourcing workers to ensure the reliability of abnormal data detection results and rationalize the remuneration of crowdsourcing workers and reduce the cost of task publishers has become a key issue in improving the crowdsourcing service of abnormal data detection.

[0003] However, most of the existing crowdsourcing worker selection methods for reliability and cost only consider the correlation between crowdsourcing worker skills and task types, and use this to ensure the reliability of crowdsourcing task results. In abnormal data detection, there is no prediction of the reliability of crowdsourcing task results by combining workers' trust values ​​or historical behaviors with worker selection schemes. Moreover, in most current studies, crowdsourcing tasks are completed by only one crowdsourcing worker. Complex and large-scale abnormal data detection tasks usually require multiple crowdsourcing workers to work together. For the case of selecting multiple crowdsourcing workers to complete a single detection task, there is a lack of a worker selection method that comprehensively considers the reliability of task results. In cost-based crowdsourcing research, most common research directions consider the moving distance of crowdsourcing workers or the consumption of batteries and traffic in mobile crowdsourcing, and there is little research on the cost of crowdsourcing task publishers, which results in the task publisher being unable to control the cost of crowdsourcing worker remuneration and the reliability of crowdsourcing task results. Summary of the invention

[0004] To this end, the present invention provides a method and system for selecting optimal workers based on reverse backpack in crowdsourcing detection of abnormal data, which comprehensively considers the expected reliability of abnormal data detection results and the bidding prices of crowdsourcing workers to select the optimal worker combination corresponding to the task requirements, thereby saving costs and improving efficiency for task publishers.

[0005] According to the design scheme provided by the present invention, an optimal worker selection method based on reverse backpack in abnormal data crowdsourcing detection includes the following contents:

[0006] Determine the set of bidding workers based on the abnormal data detection task and the crowdsourcing worker information in the crowdsourcing network; combine the crowdsourcing workers in the bidding worker set according to the requirements of the abnormal data detection task, and use the obtained multiple crowdsourcing worker combinations as candidates and add them to the candidate crowdsourcing worker set;

[0007] The trust value of crowdsourcing workers is calculated by weighted calculation to obtain the reliability of abnormal data detection result prediction. The total bid price of crowdsourcing workers is used as a constraint, the set of candidate crowdsourcing workers is used as backpack items, the trust value of crowdsourcing workers is used as the weight of backpack items, and the expected reliability threshold is set as the backpack capacity. The worker selection problem in abnormal data detection is modeled as a reverse backpack model.

[0008] The reverse knapsack model is solved, and a crowdsourcing worker combination whose predicted reliability of abnormal data detection results is greater than and / or equal to the expected reliability setting threshold is selected from the candidate crowdsourcing worker set according to the expected reliability threshold. The total bid price of the workers under this combination is used to select the optimal worker combination with the lowest total price.

[0009] As the optimal worker selection method based on reverse backpack in the abnormal data crowdsourcing detection of the present invention, further, the crowdsourcing worker set in the crowdsourcing network is traversed, and the workers in the crowdsourcing worker set who bid for the abnormal data detection task are added to the bidding worker set.

[0010] As the optimal worker selection method based on reverse backpack in the abnormal data crowdsourcing detection of the present invention, further, in the acquisition of the candidate crowdsourcing worker set, the abnormal data in the detection task is divided into m sets of data to be detected according to the generation time interval, and the bidding workers are combined and m crowdsourcing workers corresponding to the data set to be detected are selected from the bidding worker set according to the task requirements as a combination, and added to the candidate crowdsourcing worker set for abnormal detection of the data set to be detected until the candidate crowdsourcing worker set contains different worker combinations, where m is an integer and n is the number of crowdsourcing workers in the bidding worker set.

[0011] As the optimal worker selection method based on reverse backpack in the crowdsourcing detection of abnormal data in the present invention, further, by calculating the ratio of the data volume of the data set to be detected in each time interval to the total data volume of the abnormal data, and multiplying the ratio with the crowdsourcing worker trust value of the corresponding data set to be detected, the prediction reliability of the abnormal data detection result is obtained by accumulating the product of the ratio in all time intervals of the abnormal data and the crowdsourcing worker trust value.

[0012] As the optimal worker selection method based on reverse knapsack in the abnormal data crowdsourcing detection of the present invention, further, the reverse knapsack model is expressed as:

[0013] Among them, Threshold k is the expected reliability threshold of the kth anomaly data detection task, CT k is the prediction reliability of the detection result of the k-th abnormal data detection task, A flag indicating whether the i-th worker bids for the k-th abnormal data detection task, represents the bidding price of the i-th worker for the k-th abnormal data detection task, m is the number of crowdsourcing workers in each combination in the candidate crowdsourcing worker set, and n is the number of crowdsourcing workers in the bidding worker set.

[0014] As the optimal worker selection method based on reverse knapsack in abnormal data crowdsourcing detection of the present invention, the reverse knapsack model is further solved by the branch and bound method to construct a branch and bound search tree structure, in which each node represents the worker combination selection state, the number of tree layers is the number of bidding workers, and the path from the root node to the leaf node represents a group of worker selection combinations. Through all the worker combinations in the candidate crowdsourcing worker set, the optimal worker combination that meets the expected reliability threshold and has the lowest total bid price is obtained.

[0015] As the optimal worker selection method based on reverse backpack in the abnormal data crowdsourcing detection of the present invention, further, a pruning function is set according to the number of selected workers and the bid price in the solution of the reverse backpack model. When the number of selected workers meets m or the current total bid price is greater than the bid price of the currently selected worker, the pruning function is used to stop the current iterative solution process.

[0016] As the optimal worker selection method based on reverse backpack in the abnormal data crowdsourcing detection of the present invention, further, if the predicted reliability of the abnormal data detection results of all crowdsourcing worker combinations in the candidate crowdsourcing worker set is less than the expected reliability setting threshold, the expected reliability setting threshold is adjusted, and the crowdsourcing worker combination in the candidate crowdsourcing worker set is selected based on the adjusted expected reliability setting threshold.

[0017] As the optimal worker selection method based on reverse backpack in abnormal data crowdsourcing detection of the present invention, further, while outputting the selected optimal worker combination, the crowdsourcing worker trust value is updated according to the worker task execution status under the optimal worker combination.

[0018] Furthermore, the present invention also provides an optimal worker selection system based on reverse backpack in abnormal data crowdsourcing detection, comprising: a candidate combination determination module, a combination selection modeling module and a combination model solving module, wherein:

[0019] A candidate combination determination module is used to determine the set of bidding workers based on the abnormal data detection task and the crowdsourcing worker information in the crowdsourcing network; and to combine the crowdsourcing workers in the bidding worker set according to the requirements of the abnormal data detection task, and to take the obtained multiple crowdsourcing worker combinations as candidates and add them to the candidate crowdsourcing worker set;

[0020] The combined selection modeling module is used to obtain the prediction reliability of abnormal data detection results by weighted calculation of the trust value of crowdsourcing workers. The total bid price of crowdsourcing workers is used as a constraint condition, the set of candidate crowdsourcing workers is used as backpack items, the trust value of crowdsourcing workers is used as the weight of backpack items, and the expected reliability threshold is set as the backpack capacity. The worker selection problem in abnormal data detection is modeled as a reverse backpack model;

[0021] The combination model solving module is used to solve the reverse knapsack model, select a crowdsourcing worker combination whose abnormal data detection result prediction reliability is greater than and / or equal to the expected reliability setting threshold from the candidate crowdsourcing worker set according to the expected reliability threshold, and use the total bid price of the workers under this combination to select the optimal worker combination with the lowest total price.

[0022] Beneficial effects of the present invention:

[0023] The present invention considers the condition that a single crowdsourcing task corresponds to multiple crowdsourcing workers, models the worker selection problem as a reverse knapsack problem with constraints and solves it, realizes the optimal worker selection scheme for the abnormal data detection task requiring multiple workers, and more fully considers the impact of the worker selection scheme on the reliability of the abnormal data detection results and the total cost. When allocating tasks, it can give priority to meeting the requirements for the reliability of task results and minimize the total cost, thereby achieving high reliability and low cost of worker selection, and has good application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a schematic diagram of the optimal worker selection process based on reverse backpack in abnormal data crowdsourcing detection in the embodiment;

[0025] Figure 2 The invention is a schematic diagram of the principle of the optimal worker selection algorithm based on the reverse backpack in the abnormal data crowdsourcing detection in the embodiment;

[0026] Figure 3 This is a schematic diagram of the branch and bound search tree structure in the embodiment;

[0027] Figure 4 This is a schematic diagram of the crowdsourcing network structure in the embodiment;

[0028] Figure 5 This is an example of worker selection in the embodiment;

[0029] Figure 6 Schematic diagram of the crowdsourcing worker bidding and selection process in the embodiment. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solutions and advantages of the present invention clearer and more understandable, the present invention is further described in detail below in conjunction with the accompanying drawings and technical solutions.

[0031] The embodiment of the present invention provides an optimal worker selection method based on reverse backpack in abnormal data crowdsourcing detection, which includes the following contents:

[0032] S101, determining a set of bidding workers according to the abnormal data detection task and the crowdsourcing worker information in the crowdsourcing network; combining the crowdsourcing workers in the bidding worker set according to the requirements of the abnormal data detection task, and taking the obtained multiple crowdsourcing worker combinations as candidates and adding them to the candidate crowdsourcing worker set;

[0033] S102, the trust value of the crowdsourcing worker is calculated by weighted calculation to obtain the reliability of the prediction of the abnormal data detection result, the total bid price of the crowdsourcing worker is used as a constraint condition, the set of candidate crowdsourcing workers is used as backpack items, the trust value of the crowdsourcing worker is used as the weight of the backpack item, the expected reliability threshold is set as the backpack capacity, and the worker selection problem in the abnormal data detection is modeled as a reverse backpack model;

[0034] S103, solving the reverse knapsack model, selecting a crowdsourcing worker combination whose abnormal data detection result prediction reliability is greater than and / or equal to the expected reliability setting threshold from the candidate crowdsourcing worker set according to the expected reliability threshold, and using the total bid price of the workers under this combination to select the optimal worker combination with the lowest total price.

[0035] See also Figure 1 As shown in the figure, the number of crowdsourcing workers required in the detection task is taken as a constraint, the expected threshold of the reliability of the abnormal data detection results is set, and the total bid price of the crowdsourcing workers is minimized. The problem is modeled as a reverse 0-1 knapsack model with constraints, and the optimal selection of crowdsourcing workers is achieved by solving the model. The process is scientific, reasonable and easy to implement. In view of the low reliability of the abnormal data detection results and the high cost of the task publisher, the reliability of the abnormal data detection task results is evaluated using the golden data set method, while considering the bid price of the crowdsourcing workers. When selecting workers, the reliability of the abnormal data detection results is predicted by the trust value of the selected workers and the amount of data they detect. It is modeled into a reverse 0-1 knapsack problem with the total bid price of the workers and solved, thereby ensuring the reliability of the abnormal data detection results and minimizing the cost of the task publisher. This method not only considers the reliability of the abnormal data detection results and the cost of the task publisher, but also considers the number of workers required for each crowdsourcing task. In the initial stage of worker selection, the required number of workers is selected from all bidding workers and combined, in order to meet the needs of the crowdsourcing task and minimize the cost of the task publisher under the premise of ensuring the reliability of the abnormal data detection results.

[0036] As an optimal worker selection method based on reverse knapsack in the crowdsourcing detection of abnormal data in an embodiment of the present invention, further, the crowdsourcing worker set in the crowdsourcing network is traversed, and workers in the crowdsourcing worker set who bid for the abnormal data detection task are added to the bidding worker set. Furthermore, in the acquisition of the candidate crowdsourcing worker set, the abnormal data in the detection task is divided into m sets of data to be detected according to the generation time interval, and the bidding workers are traversed and combined, and m crowdsourcing workers corresponding to the data set to be detected are selected from the bidding worker set as a combination according to the task requirements, and added to the candidate crowdsourcing worker set for abnormal detection of the data set to be detected, until the candidate crowdsourcing worker set contains different worker combinations, where m is an integer and n is the number of crowdsourcing workers in the bidding worker set. Further, by calculating the ratio of the data volume of the data set to be detected in each time interval to the total data volume of the abnormal data, and multiplying the ratio with the crowdsourcing worker trust value of the corresponding data set to be detected, the prediction reliability of the abnormal data detection result is obtained by accumulating the product of the ratio in all time intervals of the abnormal data and the crowdsourcing worker trust value. Further, the branch and bound method is used to solve the reverse knapsack model, and a branch and bound search tree structure is constructed. Each node represents the worker combination selection state, the number of tree layers is the number of bidding workers, and the path from the root node to the leaf node represents a group of worker selection combinations. Through all the worker combinations in the candidate crowdsourcing worker set, the optimal worker combination that meets the expected reliability threshold and has the lowest total bidding price is obtained. Further, in the solution of the reverse knapsack model, a pruning function is set according to the number of selected workers and the bidding price. When the number of selected workers meets m or the current total bidding price is greater than the bidding price of the currently selected worker, the pruning function is used to stop the current iterative solution process. Furthermore, if the predicted reliability of the abnormal data detection results of all crowdsourcing worker combinations in the candidate crowdsourcing worker set is less than the expected reliability setting threshold, the expected reliability setting threshold is adjusted, and the crowdsourcing worker combination in the candidate crowdsourcing worker set is selected based on the adjusted expected reliability setting threshold. Furthermore, while outputting the selected optimal worker combination, the crowdsourcing worker trust value is updated based on the worker task execution status under the optimal worker combination.

[0037] See also Figure 4 As shown in / 5 / 6, suppose that in a crowdsourcing network, the task publisher publishes an abnormal data detection task, using Task k It means that the kth abnormal data detection task in the crowdsourcing network. This task requires m crowdsourcing workers, using the set Wf k = {W j ,W j+1 ,…,W j+m-1 The crowdsourcing workers bidding for this task are represented by Ta = {Ta 1 ,Ta 2 ,…,Ta n} indicates that each bidding worker Ta i Contains trust value CW i and bid price That is Ta i Contains attributes There are n workers in the crowdsourcing network, and the set W = {W 1 ,W 2 ,…W n} means that each worker W i Has its own current trust value CW iThe trust value of worker i reflects its historical behavior. The higher the trust value, the more reliable the results of abnormal data detection completed by the worker in the past. In the worker selection stage, one or more workers must be selected to complete a crowdsourcing task. Different worker selection schemes will lead to different expectations for the reliability of abnormal data detection results, and also different costs for task publishers. In order to ensure high reliability of abnormal data detection results and low cost for task publishers, the crowdsourcing network will select suitable workers to complete abnormal data detection tasks based on worker trust value and bid price when selecting workers.

[0038] The crowdsourcing task publisher publishes the abnormal data detection task in the crowdsourcing network k , all crowdsourcing workers in the crowdsourcing network, with the set W = {W 1 ,W 2 ,…,W m} indicates that the Task is initialized. k The number of crowdsourcing workers required is m. Start crowdsourcing worker selection and initialize the relevant crowdsourcing worker set. The bidding crowdsourcing worker set Meet the Task k The set of crowdsourcing worker selection sequences based on the number of crowdsourcing workers required Make Task k The set of bidding crowdsourcing workers whose expected reliability is greater than the threshold Task k Crowdsourcing worker selection results The data to be tested is divided into m smaller data sets according to the time interval in which they are generated, and m crowdsourcing workers are selected to perform anomaly detection on these data sets. k The crowdsourcing workers who bid join Ta and get the set Ta = {Ta 1 ,Ta 2 …,Ta 3}(Ta i Represents the i-th task k Workers who bid for the project). According to Task k The required number of crowdsourcing workers m performs all combinations of Ta, and adds each combination result to the sequence set Tn of workers to be selected, and obtains the sequence set {Tn 1 ,Tn 2 ,…,Tn n}(Tn i represents the combination of worker selections of the i-th type). Traverse the worker set W and add n workers who bid for this abnormal data detection task to the set Ta. Traverse the Ta set, select m workers from the Ta set, and add the selection scheme of the m workers as an element to the Tn set. Repeat this step until Tn contains According to the personal trust value of crowdsourcing workers, the reliability of the expected abnormal data detection results is calculated by a simple weighted method, and the number of workers using Task k The worker combinations whose reliability is greater than the threshold are added to Tt to obtain the sequence set {Tt 1 ,Tt 2 ,…,Tt n}. Select the group of workers with the lowest total bid price in the set Tt and add the workers in this group to the set Wf k If That is, there is no Task k The worker combination whose result reliability is greater than the threshold returns the adjusted expected reliability threshold Threshold k After the crowdsourcing worker selection is completed, the task publisher will k The trust value of the workers in this abnormal data detection is updated based on the completion of the tasks by the workers.

[0039] The meaning of each symbol is shown in the following table:

[0040]

[0041]

[0042] The reverse 0-1 knapsack problem is a traditional 0-1 knapsack model, in which Tn is analogous to the items to be put into the knapsack, and Threshold k Analogous to the capacity of a backpack, the total bid price of the selected crowdsourcing workers is analogous to the total value of the items in the backpack, the worker's personal trust value is the weight of the items in the backpack, and the objective function is to achieve the lowest total bid price of the selected crowdsourcing workers while satisfying the reliability of the predicted abnormal data detection result greater than the set threshold. The reverse 0-1 backpack problem expression is as follows:

[0043] If worker i bids for inspection task k, then The value of is 1, otherwise is 0. The branch and bound method is used to solve the reverse 0-1 knapsack problem, traverse the binary tree of bidding workers, and use the pruning function to reduce the complexity of traversal, and finally obtain the optimal worker selection solution.

[0044] See also Figure 2 As shown, the optimal worker selection algorithm for solving the reverse knapsack model can be designed as follows:

[0045]

[0046]

[0047] In the solution process, all eligible worker combinations are traversed, and a pruning function is added during the traversal process to reduce the complexity of the algorithm. Figure 3 As shown in the figure, each node represents a selection state. The number of layers of the tree is the number of workers bidding for this abnormal data detection. Each path from the root node to the leaf node represents a worker selection method. Starting from the initial state of the root node, each time it is decided whether to select the next bidding worker to reach the state of the next node. The pruning function is added during the traversal to reduce the complexity of the traversal, and the optimal worker selection scheme that meets the expected reliability threshold of the abnormal data detection results and has the lowest total bid price is obtained. Pruning function: 1. When selecting a worker, the number of currently selected workers is judged. If it meets the requirement m, the traversal is stopped. 2. When selecting a worker, if the current total price is greater than the lowest price selected by the current worker, the traversal is stopped.

[0048] Prediction of the reliability of abnormal data detection results. First, calculate the ratio of the amount of data in the data set to be tested assigned to each worker to the total amount of data to be tested, multiply the ratio by the trust value of the worker who tested the data set, and accumulate all the products to obtain the prediction result of the reliability of the abnormal data detection result when the corresponding worker selects the solution. The calculation method is shown in the following formula. It indicates the proportion of each worker's detection data to the total amount of data to be detected, so CW i is the worker’s trust value, which represents the average reliability of the abnormal data detection results completed by the worker historically, so CW i ∈[0,1].

[0049]

[0050] After the selection of crowdsourcing workers is completed, according to the quality of the crowdsourcing workers' detection of abnormal data, the trust value offset of the crowdsourcing workers participating in this task can be stored in the trust value database, and the trust value of the crowdsourcing workers can be updated. Some abnormal data are inserted into the sporadic positions of the data to be detected. These inserted data are invisible to the crowdsourcing workers. The reliability of the abnormal data detection results completed by the crowdsourcing worker is calculated by the ratio of the amount of data detected by the crowdsourcing worker in the inserted abnormal data to the amount of data inserted by the abnormal data.

[0051] It can be seen from the above technical solutions that this solution can provide a reliable and low-cost worker selection method in the abnormal data detection task where a large number of workers are required and the workers have different trust values. Among them:

[0052] Each worker selection selects the required number of workers for an abnormal data detection task, and each worker selection selects workers that provide services for only one detection task.

[0053] In each worker selection, the trust values ​​of all workers who bid for the task are first calculated to obtain the standard price and current trust value of each bidding worker.

[0054] In the first stage of worker selection, all workers who bid for the detection task will be added to the Ta set.

[0055] In the second stage of worker selection, the workers in the Ta set are combined according to the number of workers required for the detection task, and the predicted value of the reliability of the crowdsourcing task results under each group of selection options is calculated. The worker combination that makes the reliability of the abnormal data detection task results greater than the threshold is added to the Tt set. If the Tt set is empty after the above steps, return to the step before worker selection, adjust the threshold of the reliability of the abnormal data detection task results and repeat the worker selection steps. After that, the group with the lowest worker bid price is selected from the Tt set as the worker selection scheme for the task. The worker selection at this stage can be modeled as a reverse 0-1 knapsack problem, in which the worker's personal trust value is the weight of the items in the knapsack, the set threshold of the reliability of the abnormal data detection task results is the capacity of the knapsack, and the crowdsourcing worker bid price is the price of the items. The objective function is to achieve the lowest total bid price of the selected crowdsourcing workers while satisfying the reliability of the predicted abnormal data detection task results greater than the set threshold.

[0056] Finally, the crowdsourcing workers return the detected anomalies to the task publisher, who compares the anomaly data in the detection results with the gold standard set pre-inserted at sporadic locations, calculates the trust value offset of each crowdsourcing worker in the task, and publishes it in the trust database to update the trust value of the crowdsourcing worker.

[0057] To summarize, the crowdsourcing worker selection method for abnormal data detection in this case can predict the reliability of the results of this abnormal data detection task based on the reliability of the selected workers and the amount of data detected by each worker after the worker selection. The worker selection problem is then modeled and solved through the solution of this case, which can minimize the cost of the task publisher while ensuring the reliability of the abnormal data detection task results.

[0058] Furthermore, based on the above method, an embodiment of the present invention also provides an optimal worker selection system based on reverse backpack in abnormal data crowdsourcing detection, comprising: a candidate combination determination module, a combination selection modeling module and a combination model solving module, wherein:

[0059] A candidate combination determination module is used to determine the set of bidding workers based on the abnormal data detection task and the crowdsourcing worker information in the crowdsourcing network; and to combine the crowdsourcing workers in the bidding worker set according to the requirements of the abnormal data detection task, and to take the obtained multiple crowdsourcing worker combinations as candidates and add them to the candidate crowdsourcing worker set;

[0060] The combined selection modeling module is used to obtain the prediction reliability of abnormal data detection results by weighted calculation of the trust value of crowdsourcing workers. The total bid price of crowdsourcing workers is used as a constraint condition, the set of candidate crowdsourcing workers is used as backpack items, the trust value of crowdsourcing workers is used as the weight of backpack items, and the expected reliability threshold is set as the backpack capacity. The worker selection problem in abnormal data detection is modeled as a reverse backpack model;

[0061] The combination model solving module is used to solve the reverse knapsack model, select a crowdsourcing worker combination whose abnormal data detection result prediction reliability is greater than and / or equal to the expected reliability setting threshold from the candidate crowdsourcing worker set according to the expected reliability threshold, and use the total bid price of the workers under this combination to select the optimal worker combination with the lowest total price.

[0062] Unless otherwise specifically stated, the relative steps, numerical expressions and values ​​of the components and steps set forth in these embodiments do not limit the scope of the present invention.

[0063] Based on the above method and / or system, an embodiment of the present invention also provides a server, including: one or more processors; a storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the above method.

[0064] Based on the above method and / or system, an embodiment of the present invention further provides a computer-readable medium having a computer program stored thereon, wherein the above method is implemented when the program is executed by a processor.

[0065] In all examples shown and described herein, any specific values ​​should be interpreted as merely exemplary and not as limiting, and thus other examples of the exemplary embodiments may have different values.

[0066] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0067] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. An optimal worker selection method based on reverse knapsack in abnormal data crowdsourcing detection, characterized in that: Contains the following: Determine the set of bidding workers based on the abnormal data detection task and the crowdsourcing worker information in the crowdsourcing network; combine the crowdsourcing workers in the bidding worker set according to the requirements of the abnormal data detection task, and use the obtained multiple crowdsourcing worker combinations as candidates and add them to the candidate crowdsourcing worker set; The ratio of the amount of abnormal data detected by crowdsourcing workers in the inserted abnormal data to the amount of inserted abnormal data is used as the prediction reliability of the abnormal data detection result. The total bidding price of crowdsourcing workers is used as a constraint condition, the set of candidate crowdsourcing workers is used as backpack items, the trust value of crowdsourcing workers is used as the weight of backpack items, and the expected reliability threshold is set as the backpack capacity. The worker selection problem in abnormal data detection is modeled as a reverse backpack model; the reverse backpack model is expressed as: Among them, Threshold k is the expected reliability threshold of the kth anomaly data detection task, CT k is the prediction reliability of the detection result of the k-th abnormal data detection task, A flag indicating whether the i-th worker bids for the k-th abnormal data detection task, represents the bid price of the i-th worker for the k-th abnormal data detection task, m is the number of crowdsourcing workers in each combination in the candidate crowdsourcing worker set, and n is the number of crowdsourcing workers in the bidding worker set; The reverse knapsack model is solved, and a crowdsourcing worker combination whose predicted reliability of abnormal data detection results is greater than and / or equal to the expected reliability setting threshold is selected from the candidate crowdsourcing worker set according to the expected reliability threshold. The total bid price of the workers under this combination is used to select the optimal worker combination with the lowest total price.

2. The optimal worker selection method based on reverse knapsack in abnormal data crowdsourcing detection according to claim 1 is characterized in that: Traverse the crowdsourcing worker set in the crowdsourcing network, and add the workers in the crowdsourcing worker set who bid for the abnormal data detection task to the bidding worker set.

3. The optimal worker selection method based on reverse knapsack in abnormal data crowdsourcing detection according to claim 1 or 2, characterized in that: In the process of obtaining the candidate crowdsourcing worker set, the abnormal data in the detection task is divided into m sets of data to be detected according to the time interval of generation. The bidding workers are combined and m crowdsourcing workers corresponding to the data set to be detected are selected from the bidding worker set as a combination according to the task requirements. The m crowdsourcing workers are added to the candidate crowdsourcing worker set for anomaly detection of the data set to be detected until the candidate crowdsourcing worker set contains different worker combinations, where m is an integer and n is the number of crowdsourcing workers in the bidding worker set.

4. The optimal worker selection method based on reverse backpack in abnormal data crowdsourcing detection according to claim 3 is characterized in that: The prediction reliability of the abnormal data detection result is obtained by calculating the ratio of the data volume of the data set to be detected in each time interval to the total data volume of the abnormal data, and multiplying the ratio with the trust value of the crowdsourcing workers of the corresponding data set to be detected. The product of the ratio and the trust value of the crowdsourcing workers in all time intervals of the abnormal data is accumulated.

5. The optimal worker selection method based on reverse backpack in abnormal data crowdsourcing detection according to claim 1 is characterized in that: The branch and bound method is used to solve the reverse knapsack model and construct a branch and bound search tree structure. Each node represents the worker combination selection state. The number of tree layers is the number of bidding workers. The path from the root node to the leaf node represents a set of worker selection combinations. Through all the worker combinations in the candidate crowdsourcing worker set, the optimal worker combination that meets the expected reliability threshold and has the lowest total bid price is obtained.

6. The optimal worker selection method based on reverse knapsack in abnormal data crowdsourcing detection according to claim 1 or 5, characterized in that: In the solution of the reverse knapsack model, a pruning function is set according to the number of selected workers and the bid price. When the number of selected workers meets m or the current total bid price is greater than the bid price of the currently selected worker, the pruning function is used to stop the current iterative solution process.

7. The optimal worker selection method based on reverse knapsack in abnormal data crowdsourcing detection according to claim 1 is characterized in that: If the predicted reliability of the abnormal data detection results of all crowdsourcing worker combinations in the candidate crowdsourcing worker set is less than the expected reliability setting threshold, the expected reliability setting threshold is adjusted, and the crowdsourcing worker combination in the candidate crowdsourcing worker set is selected based on the adjusted expected reliability setting threshold.

8. The optimal worker selection method based on reverse knapsack in abnormal data crowdsourcing detection according to claim 1 is characterized in that: While outputting the selected optimal worker combination, the crowdsourcing worker trust value is updated according to the worker task execution status under the optimal worker combination.

9. An optimal worker selection system based on reverse backpack in abnormal data crowdsourcing detection, characterized in that: It includes: candidate combination determination module, combination selection modeling module and combination model solving module, among which, A candidate combination determination module is used to determine the set of bidding workers based on the abnormal data detection task and the crowdsourcing worker information in the crowdsourcing network; and to combine the crowdsourcing workers in the bidding worker set according to the requirements of the abnormal data detection task, and to take the obtained multiple crowdsourcing worker combinations as candidates and add them to the candidate crowdsourcing worker set; The combined selection modeling module is used to take the ratio of the amount of abnormal data detected by the crowdsourcing workers in the inserted abnormal data to the amount of inserted abnormal data as the reliability of the abnormal data detection result prediction, use the total bid price of the crowdsourcing workers as the constraint condition, the candidate crowdsourcing workers set as the backpack items, the crowdsourcing workers' trust value as the backpack item weight, and the expected reliability threshold as the backpack capacity. The worker selection problem in abnormal data detection is modeled as a reverse backpack model; wherein the reverse backpack model is expressed as: Among them, Threshold k is the expected reliability threshold of the kth anomaly data detection task, CT k is the prediction reliability of the detection result of the k-th abnormal data detection task, A flag indicating whether the i-th worker bids for the k-th abnormal data detection task, represents the bid price of the i-th worker for the k-th abnormal data detection task, m is the number of crowdsourcing workers in each combination in the candidate crowdsourcing worker set, and n is the number of crowdsourcing workers in the bidding worker set; The combination model solving module is used to solve the reverse knapsack model, select a crowdsourcing worker combination whose abnormal data detection result prediction reliability is greater than and / or equal to the expected reliability setting threshold from the candidate crowdsourcing worker set according to the expected reliability threshold, and use the total bid price of the workers under this combination to select the optimal worker combination with the lowest total price.

Images