A method and system for selecting participants in crowd intelligence perception based on prediction of cooperation willingness

Abstract

The present invention relates to the field of crowd intelligence sensing technology, and in particular to a crowd intelligence sensing participant selection method and system based on cooperation willingness prediction. It includes obtaining a participant data set, obtaining an estimated cooperation cost through platform tasks; constructing a multi-armed bandit problem model, virtually employing the participant data set, and calculating the participant pair exploration rate; setting a threshold for the participant pair exploration rate, processing the data set using a link prediction model to obtain a predicted cooperation cost, and obtaining a converged cooperation cost based on the employed worker pair data and the estimated cooperation cost or predicted cooperation cost; obtaining a participant set using a minimum cost priority participant selection strategy; and updating the estimated cooperation cost and the employed worker pair data. It solves the problem of how the platform selects participants to minimize the cost of task completion when the relationship between the participants is not completely known or there is no historical data of a large number of participant pairs cooperating.

Description

Technical Field

[0001] The present invention relates to the field of crowd sensing technology, and in particular to a crowd sensing participant selection method and system based on cooperation willingness prediction. Background Art

[0002] With the advancement of mobile communications, sensor technology and chip technology, smart portable terminals, such as smartphones and tablets, include global positioning systems (GPS), accelerometers, thermometers, electronic compasses and gyroscopes, making smart portable terminals no longer a single communication device. A large number of smart portable terminals are distributed in different geographical locations, which can easily collect environmental information and connect to network transmission, thus giving rise to a group intelligent perception system. Group intelligence perception provides a forward-looking solution for large-scale information collection. Group intelligence perception is a collaborative system that requires recruiting participants to complete multiple tasks.

[0003] In previous research on crowd intelligence perception, most of the focus was on how to reasonably employ participants in crowd intelligence perception task scenarios that can be completed by one person to achieve lower task completion costs. In reality, there are still many complex tasks that require cooperation among multiple people. Such scenarios can be called atomic cooperative task scenarios, which require multiple participants to achieve through good communication and cooperation. Obviously, communication between participants requires costs. The closer the relationship between participants, that is, the stronger the willingness to cooperate, the lower the cooperation cost between participants. On the one hand, crowd intelligence perception is a "people-centered" perception. Ignoring the user's willingness to cooperate may lead to low user execution efficiency or even quitting midway, refusing to cooperate, etc. Therefore, the willingness of participants to participate is its subjective factor; on the other hand, the trust relationship between members plays an important role in team operation. It can reduce the risk and uncertainty of cooperation. Trust between users is crucial to ensure the success and quality of cooperation. Having a high degree of trust is an objective factor that the platform hopes to form a team. The existing technology lacks the construction of a collaborative team by considering the willingness relationship and trust relationship between users at the same time, which leads to poor service quality. The platform needs to think about how to reasonably employ participants to achieve lower task completion costs while ensuring the task completion rate.

[0004] In the prior art, there have been some studies in cooperative task scenarios, which use the social network information of participants to judge the relationship between participants and combine the capabilities of participants to make reasonable participant hiring decisions. However, with the development of society, the importance of personal privacy of participants' social networks has increased, and it is obviously very difficult to obtain the social network data of all participants. How to make reasonable participant hiring decisions when the relationship between participants is not fully known or there is no historical data of a large number of participants' cooperation is another new challenge brought about by atomic cooperative task scenarios. Summary of the invention

[0005] In order to solve the above-mentioned problem of how the platform selects participants to minimize the cost of task completion when the relationship between participants is not completely known or there is no historical data of cooperation among a large number of participants, the present invention provides a method and system for selecting participants based on prediction of cooperation willingness.

[0006] In a first aspect, the present invention provides a method for selecting participants in crowd intelligence perception based on cooperation willingness prediction, which adopts the following technical solution:

[0007] A method for selecting participants in crowd sensing based on prediction of cooperation willingness includes:

[0008] Obtain participant data sets and obtain estimated collaboration costs through platform tasks;

[0009] Construct a multi-armed bandit problem model, perform virtual employment on the participant data set, record the number of times the participant is employed, obtain the data of the employed worker pairs, and calculate the participant pair exploration rate;

[0010] A threshold of the exploration rate of the participants is set. When the exploration rate of the participants is not less than the threshold, the data set is processed using a link prediction model to obtain a predicted cooperation cost. The converged cooperation cost is obtained based on the hired workers' data and the predicted cooperation cost. When the exploration rate of the participants is less than the threshold, the converged cooperation cost is obtained based on the hired workers' data and the estimated cooperation cost.

[0011] The participant set is obtained by using the minimum cost first participant selection strategy;

[0012] Updates to the estimated collaboration costs and worker pair data.

[0013] Furthermore, the estimated cooperation cost is the average of the estimated maximum cooperation cost and the minimum cooperation cost, and the estimated cooperation cost is expressed in a matrix as follows: ,in, represents the estimated cooperation cost matrix, represents the minimum cooperation cost, represents the maximum cooperation cost; Represents the number of participants in the participant dataset.

[0014] Furthermore, the data of the employed workers is a count matrix ,in represents the number of participants in the participant dataset, The round variable representing the task, Indicates the initialization round.

[0015] Furthermore, the participant pair exploration rate is a participant pair consisting of two participants, and the participant pair exploration rate is expressed by the formula:

[0016]

[0017] in, is the exploration rate of participants, which is used to indicate the proportion of participants whose number of employments is greater than 1. Represents the participants in the count matrix and The number of times the participant pairs were hired.

[0018] Furthermore, the threshold setting range of the exploration rate is , the exploration rate threshold is expressed as .

[0019] Furthermore, the processing of the data set using the link prediction model includes:

[0020] Obtaining data of employed workers as a training set, and using the training set to train the link prediction model;

[0021] Inputting the participant data set into a trained link prediction model to obtain a prediction result;

[0022] Organize the estimated cooperation costs of all participants into a predicted cooperation cost matrix .

[0023] Furthermore, the method of obtaining the converged cooperation cost based on the data and predicted cooperation cost of the employed workers includes calculating the upper confidence bound index of the cooperation cost matrix to obtain the converged cooperation cost matrix, which is expressed as follows:

[0024]

[0025] in, represents the calculated participant pair ( ) is the upper confidence bound index of the cooperation cost, Indicates the current pair of participants ( ) of the estimated cost of cooperation, represents the adjustment parameters, is a round variable, which is used to indicate the current task round. represents the forecast cooperation cost matrix.

[0026] Furthermore, the method of obtaining the converged cooperation cost based on the data of the employed workers and the estimated cooperation cost includes calculating the upper confidence bound index of the cooperation cost matrix to obtain the converged cooperation cost matrix, which is expressed as follows:

[0027] .

[0028] Furthermore, the participant selection strategy using minimum cost first is used to obtain the participant set, including:

[0029] Create and initialize a collection , get the workers' execution quotation set for data and platform tasks , select the participant with the lowest bid to join the set ;

[0030] Using execution quote collection and convergence cooperation cost matrix to loop and select participants to join the set , including the participant whose sum of cooperation cost and execution bid is the smallest each time;

[0031] Repeat the selection until the set The number of participants in reaches the number of participants required by the task.

[0032] Further, the updating of the estimated collaboration cost and the data of the employed workers includes utilizing the set The participants in perform tasks and obtain the actual cooperation cost of each worker pair, and update the matrix of estimated cooperation cost. The formula is:

[0033]

[0034] in, is the estimated cooperation cost matrix The elements in are represented by the participants before updating ( ) of the estimated cost of cooperation, The round variable representing the task, Indicates that in executing In the second task, participants )’s actual cost of cooperation,

[0035] is the count matrix The elements in are the participants' pairs ( ) in the technology matrix, indicating the participants' ) were employed, Representing a collection Any two different participants and ,

[0036] Update the count matrix, the formula is: .

[0037] Second, a crowd-sensing participant selection system based on cooperation willingness prediction includes:

[0038] The data acquisition module is configured to acquire the data set of participants and obtain the estimated cooperation cost through the platform tasks;

[0039] The employment simulation module is configured to construct a multi-armed bandit problem model, perform virtual employment on the participant data set, record the number of times the participant is employed, obtain the data of the employed worker pairs, and calculate the exploration rate of the participant pairs;

[0040] The converged cooperation cost determination module is configured to set a threshold value of the exploration rate of the participant pair, and when the exploration rate of the participant pair is not less than the threshold value, use the link prediction model to process the data set to obtain the predicted cooperation cost, and obtain the converged cooperation cost according to the hired workers' pair data and the predicted cooperation cost; when the exploration rate of the participant pair is less than the threshold value, obtain the converged cooperation cost according to the hired workers' pair data and the estimated cooperation cost;

[0041] The participant selection module is configured to obtain a participant set using a minimum cost first participant selection strategy;

[0042] The data updating module is configured to update the estimated cooperation cost and the hired worker pair data.

[0043] In a third aspect, the present invention provides a computer-readable storage medium storing a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device, a method for selecting participants of a crowd intelligence perception based on prediction of cooperation willingness.

[0044] In a fourth aspect, the present invention provides a terminal device comprising a processor and a computer-readable storage medium, the processor being used to implement various instructions; the computer-readable storage medium being used to store a plurality of instructions, the instructions being suitable for being loaded by the processor and executing the method for selecting participants of a crowd intelligence perception based on prediction of cooperation willingness.

[0045] In summary, the present invention has the following beneficial technical effects:

[0046] 1. The present invention proposes a method and system for selecting participants based on the prediction of cooperation willingness, introduces a multi-armed bandit problem model, performs virtual employment on the participant data set, records the number of employments and calculates the exploration rate, provides a dynamic evaluation mechanism for participant selection, and adapts to the needs of making decisions when the participant relationship is not fully known or the historical data is insufficient. In combination with the link prediction model, the data set is processed by link prediction, and the data of the employed worker pairs and the predicted cooperation cost or the estimated cooperation cost are combined according to different exploration rates to obtain the converged cooperation cost. When selecting the participant pairs, the estimated value and uncertainty of the cooperation cost are comprehensively considered, making the decision more reasonable.

[0047] 2. The present invention takes into account the willingness to cooperate and the trust relationship. In the process of participant selection, it not only focuses on the cost of task completion, but also considers subjective and objective factors such as the willingness to cooperate and the trust relationship between participants, thereby improving the execution efficiency of participants and reducing the situation of dropping out and refusing to cooperate. By utilizing the minimum cost-first participant selection strategy, it is possible to reasonably employ participants while ensuring the task completion rate, achieve a lower task completion cost, and provide an economical and efficient solution for the platform. By continuously updating the estimated cooperation cost and the data of the employed workers, the system can be continuously optimized and adjusted as the task progresses, improving the accuracy and adaptability of decision-making, and providing strong support for long-term crowd intelligence perception projects.

[0048] 3. The method of the present invention can adapt to different data states and make reasonable participant employment decisions when the relationship between participants is not completely known or there is no large amount of historical data. The present invention has been verified through simulation experiments. The method proposed by the present invention can reduce the total cost of task completion by 1.57%, and the possibility of the platform selecting participants with cooperation willingness is increased by up to 19.5%, thereby improving the economic benefits of the platform. The method is effective and superior in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 This is a scene example diagram used in the present invention.

[0050] Figure 2 Select a step diagram for the participants of the present invention.

[0051] Figure 3 This is a schematic diagram of the relationship between the number of participants and the total cost of completing a task when the method described in the present invention is used.

[0052] Figure 4 Schematic diagram of the relationship between the total cost of completing a task and the difference between BWS and the number of task rounds when the method of the present invention is used.

[0053] Figure 5is the threshold value of the link prediction intervention exploration rate when using the method of the present invention Schematic diagram of the relationship with the number of task rounds.

[0054] Figure 6 The figure is a schematic diagram showing the relationship between the hit probability of finding a pair of participants with cooperation willingness and the number of task rounds when the method of the present invention is used. DETAILED DESCRIPTION

[0055] The present invention is further described in detail below in conjunction with the accompanying drawings.

[0056] Example 1

[0057] Reference Figure 1 ,The crowd intelligence perception scenario includes three roles: requester, platform and participant. Requester refers to individuals, companies, organizations or government agencies with data collection needs. Platform refers to the cloud platform responsible for convening and organizing participants to complete services for requesters, and participants are people who are willing to perform tasks. Figure 2 The interaction process between the three roles during the execution of the crowd intelligence perception task is as follows: 1. The requester sends an atomic cooperation task request to the platform; 2. The platform publishes task information to the participant group; 3. The participant returns the task quotation to the platform; 4. The platform selects and employs a set of participants to cooperate in the execution of the atomic cooperation task according to the method of this application; 5. The participant set returns the task result to the platform, which is collected and processed by the platform and returned to the requester; 6. The requester pays the platform, and the platform pays the requester.

[0058] Reference Figure 2 , a method for selecting a participant of crowd intelligence perception based on cooperation willingness prediction in this embodiment includes:

[0059] S1. Obtain the participant data set and obtain the estimated cooperation cost through platform tasks;

[0060] The estimated cooperation cost is the average of the estimated maximum cooperation cost and the estimated minimum cooperation cost. The estimated cooperation cost is expressed in a matrix as follows: ,in, represents the estimated cooperation cost matrix, represents the minimum cooperation cost, represents the maximum cooperation cost; Represents the number of participants in the participant dataset.

[0061] Relevant information of participants is collected from the database of the crowd intelligence perception platform, including but not limited to the participants' skill level, historical task performance, willingness to participate, etc. This information constitutes the participant data set.

[0062] The estimation method for determining the minimum cooperation cost and the maximum cooperation cost can be achieved by analyzing the range of cooperation costs of similar participant combinations in historical tasks, or by estimating based on experience.

[0063] S2. Construct a multi-armed bandit problem model, perform virtual employment on the participant data set, record the number of times the participant is employed, obtain the data of the employed worker pairs, and calculate the participant pair exploration rate;

[0064] In this embodiment, it is critical to introduce the multi-armed bandit problem model. Since the cooperation cost between participants is unknown and fluctuating, and the platform needs to consider the unknown cooperation cost to select a set of participants to perform tasks, this method is modeled as a multi-armed bandit problem in order to solve the problem of online selection of participants with unknown cooperation costs.

[0065] The method models the participant selection problem as a multi-armed combined multi-armed bandit problem. In this problem model, a participant pair formed by the combination of two participants is regarded as a simple arm. The combined arm is a participant set that meets the requirements of the atomic cooperation task. The attribute learned by the arm is the cooperation cost of the participant combination. ,The reward is the total cost of the participants to complete the task. Based on the upper confidence bound (UCB) method, the cooperation cost matrix between participants is estimated before selecting participants. Each item , add an adjustment To balance the number of exploration and utilization of different participant combinations, called the UCB index .

[0066] S21. The data of the employed workers is a count matrix ,in represents the number of participants in the participant dataset, The round variable representing the task, Indicates the initialization round.

[0067] S22. The participant pair exploration rate is a participant pair consisting of two participants, and the participant pair exploration rate is expressed by the formula: ,

[0068] in, is the exploration rate of participants, which is used to indicate the proportion of participants whose number of employments is greater than 1. Represents the participants in the count matrix and The number of times the participant pairs were hired.

[0069] According to actual conditions and experience, the threshold setting range of the exploration rate is , the exploration rate threshold is expressed as .

[0070] Theoretically, the threshold setting range of exploration rate is However, under a smaller threshold, the number of participants’ willingness to cooperate explored is too small to build the training set and test set required for training the link prediction model, resulting in training failure. Under a larger threshold, the number of participants’ willingness to cooperate explored by the platform is close to the total number of participants’ willingness to cooperate, which makes it meaningless to use the model to predict the participants’ willingness to cooperate, and has no beneficial effect on the exploration process. Through the analysis of the experimental results, we found that the threshold setting in [0.3, 0.7) has a better effect.

[0071] By simulating the hiring process, the number of times a participant is hired is recorded, and the exploration rate is calculated. The exploration rate reflects the degree of exploration of different combinations of participants by the platform, helping the platform determine whether it needs to further explore new combinations of participants. According to the size of the exploration rate, different methods are selected to process the data set to obtain the converged cooperation cost. The converged cooperation cost is a more accurate estimate of the cooperation cost between pairs of participants.

[0072] S3. Set a threshold for the exploration rate of the participants. When the exploration rate of the participants is not less than the threshold, use the link prediction model to process the data set to obtain the predicted cooperation cost, and obtain the converged cooperation cost based on the hired workers' data and the predicted cooperation cost. When the exploration rate of the participants is less than the threshold, obtain the converged cooperation cost based on the hired workers' data and the estimated cooperation cost;

[0073] The step S31 of processing the data set using the link prediction model includes:

[0074] S311. Obtaining data of employed workers as a training set, and using the training set to train the link prediction model;

[0075] S312. Input the participant data set into the trained link prediction model to obtain a prediction result;

[0076] S313. Organize the estimated values ​​of cooperation costs of all participants into a predicted cooperation cost matrix .

[0077] In this embodiment, the SEAL link prediction method is used to predict the existence of links. SEAL is a link prediction method based on graph neural network, which learns a mapping function from subgraph to link existence suitable for the current network by sampling some connections and non-connections in the network structure data as sample labels and extracting subgraphs around the labels as feature data.

[0078] Specific steps for using link prediction method to predict cooperation costs:

[0079] (1) Build a training set for link prediction. The platform maintains a 0-1 matrix of links that have been explored , Any element in The value can be 1 or 0. The meaning is:

[0080]

[0081] Each time a group of participants is hired to complete a task When a task is executed, a record is returned. Matrix of communication costs between participants in the process , for the location of the employed participant, , while the unemployed participants position, . Set a threshold ,when When , it is considered that the participants have the willingness to cooperate and have links. The value of the corresponding position Otherwise, when , it means that there is no willingness to cooperate between the participants, that is, there is no link. After that, the constructed matrix It can be used to train the link prediction method. That is, after each task is completed, the cost matrix Then, the link matrix Update as follows:

[0082]

[0083] Will get The matrix is ​​used as the training set to train the link prediction model. After the training is completed, the trained model is represented as .

[0084] (2) Construct a test set and use the trained model to make predictions. , find all unexplored participant relationships, all unexplored participant relationships are represented by the set It is represented by the count matrix The position of the elements that are equal to 1 in .

[0085]

[0086] After getting the test set, test the set Input to the trained model ,get The prediction result set of the middle position .

[0087]

[0088] (3) Based on the link prediction results, build the participants’ prediction of the cooperation cost. First, according to the matrix Initialize a prediction cooperation cost matrix with the value of ,according to The prediction results will be The participant pairs with a median value of 1 are predicted to have a willingness to cooperate. The value of the corresponding position in is assigned to .

[0089]

[0090] The key to using the link prediction model in this embodiment is that since the multi-armed bandit model needs to learn a large number of participant pairs with a large number of cooperation costs (the square of the number of participants), it takes many rounds of exploration to explore participants with a willingness to cooperate. In order to speed up the process and reduce the cost of trial and error, a link prediction method is introduced. During the exploration process, the link prediction model is conditionally trained and predictions are made. For the participant pairs predicted to be willing to cooperate, priority exploration is performed at the appropriate time. In this way, when the platform explores and hires participants for tasks, it increases the probability of hiring participant pairs with a willingness to cooperate, reduces the trial and error cost of exploration, and thus reduces the total task execution cost.

[0091] S32. The method of obtaining the converged cooperation cost based on the data and predicted cooperation cost of the employed workers includes calculating the upper confidence bound index of the cooperation cost matrix to obtain the converged cooperation cost matrix, which is expressed as follows:

[0092] ,

[0093] in, represents the calculated participant pair ( ) of the cooperation cost upper confidence bound index (UCB index), Indicates the current pair of participants ( ) of the estimated cost of cooperation, represents the adjustment parameters, is a round variable, which is used to indicate the current task round. represents the forecast cooperation cost matrix.

[0094] S33. The method of obtaining the converged cooperation cost based on the data and the estimated cooperation cost of the employed workers includes calculating the upper confidence bound index of the cooperation cost matrix to obtain the converged cooperation cost matrix, which is expressed as follows:

[0095] .

[0096] S4. Use the minimum cost first participant selection strategy to obtain the participant set, that is, to obtain the participants selected for employment to perform the task.

[0097] The participant selection strategy using the minimum cost first approach to obtain the participant set includes:

[0098] S41. Create and initialize a collection , get the workers' execution quotation set for data and platform tasks , select the participant with the lowest bid to join the set ;

[0099] S42. According to the task requirements and using the execution quotation collection and convergence cooperation cost matrix to loop and select participants to join the set , including the participant whose sum of cooperation cost and execution bid is the smallest each time;

[0100] S43. Repeat the selection until the set The number of participants in reaches the number of participants required by the task.

[0101] S5. Update the estimated collaboration costs and the data on the number of workers hired.

[0102] S51. The updating of the estimated collaboration cost and the data of the workers employed includes utilizing the collection The participants in perform tasks and obtain the actual cooperation cost of each worker pair, and update the matrix of estimated cooperation cost. The formula is:

[0103] ,

[0104] in, is the estimated cooperation cost matrix The elements in are represented by the participants before updating ( ) of the estimated cost of cooperation, The round variable representing the task, Indicates that in executing In the second task, participants )’s actual cost of cooperation,

[0105] is the count matrix The elements in are the participants' pairs ( ) in the technology matrix, indicating the participants' ) were employed, Representing a collection Any two different participants and ,

[0106] S52. Update the count matrix, the formula is: .

[0107] The participant set is obtained by using the minimum cost priority participant selection strategy, and a set of participants with the most cost advantage is selected through a series of steps to perform the task. This process comprehensively considers the execution bids of the participants and the predicted value of the cooperation cost to achieve the goal of minimizing the cost of completing the task.

[0108] The cooperation cost matrix is ​​updated. As the tasks are performed, the actual cooperation costs are gradually revealed. By taking the actual cooperation costs into account and updating the cooperation cost matrix, the estimates of the cooperation costs between different pairs of participants can be more accurate. The initial cooperation cost estimate may be just a broad estimate based on the average of the maximum and minimum cooperation costs. As the actual data continues to accumulate, the updated estimate will gradually approach the true cooperation cost. The cooperation costs between participants may change with time, task type, environment and other factors. Continuously updating the cooperation cost matrix allows the platform to better adapt to these dynamic changes. Accurate cooperation cost estimates are crucial to selecting the optimal set of participants. By continuously updating the cooperation cost matrix, the platform can make more informed decisions in each task to minimize the cost of task completion.

[0109] Update the count matrix, which records the number of times a participant pair is hired, and the calculation of the exploration rate depends on this count matrix. By updating the count matrix, the number of participant pairs that have been hired more than 1 can be accurately counted, so as to calculate a more accurate exploration rate. The count matrix can intuitively reflect which participant pairs are frequently used and which participant pairs have not been fully explored. This helps the platform understand the distribution of cooperation between participants and provides a basis for balancing the use of different participant pairs. In long-term crowd perception tasks, the update of the count matrix can provide historical data references for the platform. Over time, the platform can analyze the changing trends of the count matrix to understand the stability and changing laws of cooperation between participants, so as to make more forward-looking decisions.

[0110] In order to evaluate the effect of the method of the present invention in reducing the cost of task completion, we conducted simulation experiments based on a real-world collaborator network dataset under different problem scales and parameters. We also introduced three comparison methods, namely BWS (Basic Worker Selection), CWS (CMAB-based Worker Selection), and LPWS (Link Prediction based Worker Selection), to analyze and compare with the method proposed in this embodiment (expressed as WSCLP, Worker Selection based on CMAB and Link Prediction).

[0111] Reference Figure 3 , a comparison of the total cost of task completion when the crowd intelligence perception platform adopts four participant selection methods under different numbers of participants is given. As the number of candidate participants increases, the total cost of task completion of the four schemes increases. This is because the increase in the number of participants increases the number of participant pairs that the platform can choose, but the proportion of participant pairs with cooperative willingness gradually decreases, which reduces the probability of the platform selecting participant pairs with cooperative willingness and increases the trial and error cost. Under different numbers of participants, the method of this application (WSCLP) can always achieve the lowest total cost of task completion, which can reduce the total cost of task completion by up to 1.57% compared with the BWS method.

[0112] Reference Figure 4 , the difference between the total cost of task completion when the crowd intelligence perception platform adopts four participant selection methods and the total cost of task completion of the BWS method under different task rounds is given. When the number of task rounds is about 1200-4000, the total cost of task completion of the methods WSCLP and LPWS of the present invention is higher than that of BWS. This is because the number of participant pairs with cooperative willingness accumulated in the early stage is small, and the link prediction method has not obtained a good training effect. In the later stage (after 4000 rounds), the total cost of task completion of the method of the present invention is significantly lower than that of other comparison methods.

[0113] Reference Figure 5 , given at different link prediction intervention thresholds Comparison of the total cost of completing the task when the crowd intelligence perception platform uses four participant selection methods. When the total cost of completing the task is between 0.3 and 0.5, the method of the present invention can achieve the lowest total cost of completing the task. When it is greater than 0.5, the method of the present invention is slightly worse than the CWS method but better than other methods.

[0114] Reference Figure 6 , the hit rate comparison of the group intelligence perception platform exploring the participants with cooperation willingness when using four participant selection methods under different task rounds is given. It can be seen from the figure that from 0 to 1400 rounds, the hit rates of the four methods are not much different. After 1400 rounds, the hit rates of the methods WSCLP and LPWS of the present invention are significantly improved. This is because the link prediction model intervenes in the participant selection process, which can verify the role of the link prediction model in the participant selection method. As the number of rounds increases, the hit rate of the method of the present invention is increased by 19.5% compared with BWS and CWS. After all task rounds are completed, the final hit rate is 6.4% higher than that of BWS.

[0115] Example 2

[0116] This embodiment provides a crowd-sensing participant selection system based on cooperation willingness prediction, including:

[0117] The data acquisition module is configured to acquire the data set of participants and obtain the estimated cooperation cost through the platform tasks;

[0118] The employment simulation module is configured to construct a multi-armed bandit problem model, perform virtual employment on the participant data set, record the number of times the participant is employed, obtain the data of the employed worker pairs, and calculate the exploration rate of the participant pairs;

[0119] The converged cooperation cost determination module is configured to set a threshold value of the exploration rate of the participant pair, and when the exploration rate of the participant pair is not less than the threshold value, use the link prediction model to process the data set to obtain the predicted cooperation cost, and obtain the converged cooperation cost according to the hired workers' pair data and the predicted cooperation cost; when the exploration rate of the participant pair is less than the threshold value, obtain the converged cooperation cost according to the hired workers' pair data and the estimated cooperation cost;

[0120] The participant selection module is configured to obtain a participant set using a minimum cost first participant selection strategy;

[0121] The data updating module is configured to update the estimated cooperation cost and the hired worker pair data.

[0122] A computer-readable storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device for a method of selecting a participant of a crowd intelligence perception based on prediction of cooperation willingness.

[0123] A terminal device includes a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, wherein the instructions are suitable for being loaded by the processor and executing the method for selecting participants of crowd intelligence perception based on prediction of cooperation willingness.

[0124] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for selecting participants in crowd intelligence perception based on prediction of cooperation willingness, characterized in that: include: Obtain participant data sets and obtain estimated collaboration costs through platform tasks; Construct a multi-armed bandit problem model, perform virtual employment on the participant data set, record the number of times the participant is employed, obtain the data of the employed worker pairs, and calculate the participant pair exploration rate. The participant pair exploration rate is a participant pair consisting of two participants. The participant pair exploration rate is expressed by the formula: ,in, is the exploration rate of participants, represents the count matrix of the data of the workers hired, Represents the participants in the count matrix and The number of times the participants in the composition were employed; Set the exploration rate threshold of the participant, and the exploration rate threshold setting range is ,When the exploration rate of the participants is not less than the threshold, the link prediction model is used to process the data set to obtain the predicted cooperation cost, and the converged cooperation cost is obtained according to the hired workers' data and the predicted cooperation cost. When the exploration rate of the participants is less than the threshold, the converged cooperation cost is obtained according to the hired workers' data and the estimated cooperation cost. The converged cooperation cost includes calculating the upper confidence bound index of the cooperation cost matrix, and obtaining the converged cooperation cost matrix, which is expressed as follows: ,in, represents the upper confidence bound index of the calculated cooperation cost of the participant pair, represents the current estimate of the cooperation cost for the participant pair, represents the adjustment parameters, is the round variable, represents the forecast cooperation cost matrix, represents the estimated collaboration cost matrix; The participant set is obtained by using the minimum cost first participant selection strategy; Updates to the estimated collaboration costs and worker pair data.

2. According to claim 1, a method for selecting participants in crowd intelligence perception based on cooperation willingness prediction is characterized in that: The method of processing the data set by using the link prediction model includes arranging the estimated values ​​of the cooperation costs of all pairs of participants into a predicted cooperation cost matrix by using the link prediction model.

3. According to claim 1, a method for selecting participants in crowd intelligence perception based on cooperation willingness prediction is characterized in that: The participant selection strategy using the minimum cost first approach to obtain the participant set includes: Create and initialize a collection , obtain the set of workers' execution bids for data-based platform tasks, and select the participant with the lowest bid to join the set ; Using the execution quotation set and the converged cooperation cost matrix, participants are selected to join the set in a loop , including the participant whose sum of cooperation cost and execution bid is the smallest each time; Repeat the selection until the set The number of participants in reaches the number of participants required by the task.

4. According to claim 1, a method for selecting participants in crowd intelligence perception based on prediction of cooperation willingness is characterized in that: The update estimates the cost of collaboration and the data on the workers employed, including the use of the collection The participants in perform the task and obtain the actual cooperation cost of each worker pair, update the matrix of estimated cooperation cost with the actual cooperation cost, and update the count matrix.

5. A crowd-sensing participant selection system based on cooperation willingness prediction, characterized in that: include: The data acquisition module is configured to acquire the data set of participants and obtain the estimated cooperation cost through the platform tasks; The employment simulation module is configured to construct a multi-armed bandit problem model, perform virtual employment on the participant data set, record the number of times the participant is employed, obtain the data of the employed worker pairs, and calculate the participant pair exploration rate. The participant pair exploration rate is a participant pair consisting of two participants. The participant pair exploration rate is expressed by the formula: ,in, is the exploration rate of participants, represents the count matrix of the data of the workers hired, Represents the participants in the count matrix and The number of times the participants in the composition were employed; The convergence cooperation cost determination module is configured to set a threshold value of the exploration rate of the participant, and the threshold setting range of the exploration rate is ,When the exploration rate of the participants is not less than the threshold, the link prediction model is used to process the data set to obtain the predicted cooperation cost, and the converged cooperation cost is obtained according to the hired workers' data and the predicted cooperation cost. When the exploration rate of the participants is less than the threshold, the converged cooperation cost is obtained according to the hired workers' data and the estimated cooperation cost. The converged cooperation cost includes calculating the upper confidence bound index of the cooperation cost matrix, and obtaining the converged cooperation cost matrix, which is expressed as follows: ,in, represents the upper confidence bound index of the calculated cooperation cost of the participant pair, represents the current estimate of the cooperation cost for the participant pair, represents the adjustment parameters, is the round variable, represents the forecast cooperation cost matrix, represents the estimated collaboration cost matrix; The participant selection module is configured to obtain a participant set using a minimum cost first participant selection strategy; The data updating module is configured to update the estimated cooperation cost and the hired worker pair data.

6. A computer-readable storage medium storing a plurality of instructions, characterized in that: The instructions are suitable for being loaded by a processor of a terminal device and executing a method for selecting crowd intelligence perception participants based on prediction of cooperation willingness as described in claim 1.

7. A terminal device, comprising a processor and a computer-readable storage medium, wherein the processor is used to implement each instruction; and the computer-readable storage medium is used to store multiple instructions, characterized in that: The instructions are suitable for being loaded by a processor and executing a method for selecting crowd intelligence perception participants based on prediction of cooperation willingness as described in claim 1.

Images