CN114372680A - Spatial crowdsourcing task allocation method based on worker loss prediction - Google Patents
Spatial crowdsourcing task allocation method based on worker loss prediction Download PDFInfo
- Publication number
- CN114372680A CN114372680A CN202111589692.4A CN202111589692A CN114372680A CN 114372680 A CN114372680 A CN 114372680A CN 202111589692 A CN202111589692 A CN 202111589692A CN 114372680 A CN114372680 A CN 114372680A
- Authority
- CN
- China
- Prior art keywords
- worker
- task
- satisfaction
- workers
- loss
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 37
- 239000013598 vector Substances 0.000 claims description 18
- 230000006870 function Effects 0.000 claims description 11
- 230000008447 perception Effects 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 8
- 230000004913 activation Effects 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000001914 filtration Methods 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 4
- 230000003044 adaptive effect Effects 0.000 claims description 2
- 230000008859 change Effects 0.000 claims description 2
- 239000011159 matrix material Substances 0.000 claims description 2
- 238000005457 optimization Methods 0.000 abstract description 2
- 230000006399 behavior Effects 0.000 description 7
- 238000013528 artificial neural network Methods 0.000 description 6
- 238000011160 research Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 230000002596 correlated effect Effects 0.000 description 2
- 230000000875 corresponding effect Effects 0.000 description 2
- 238000012417 linear regression Methods 0.000 description 2
- 238000012358 sourcing Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
- G06Q10/06311—Scheduling, planning or task assignment for a person or group
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Educational Administration (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a space crowdsourcing task allocation method based on worker loss prediction. The method comprises the steps that implicit feelings of workers are extracted by using an LSTM model, idle time intervals of the workers are learned from execution data of historical tasks of the workers, the model captures satisfaction and enthusiasm of the workers through two paths of LSTMs, then predicted idle time interval tensors are output through a full connection layer, and the easy-to-lose condition of the workers is judged according to a time interval threshold; and then, according to a greedy algorithm sorted by idle time, adding associated edge number limitation and combining with an actual optimal weight matching task allocation algorithm of an optimization increasing strategy, the efficiency and effectiveness of task allocation in spatial crowdsourcing are improved, the experience of workers prone to loss in the spatial crowdsourcing problem is improved, and accordingly worker loss is reduced.
Description
Technical Field
The invention relates to a data transmission method technology in communication, in particular to a spatial crowdsourcing task allocation method based on worker loss prediction.
Background
With the popularization of network devices equipped with GPS, such as smartphones, a new type of crowdsourcing, namely Spatial Crowdsourcing (SC), has attracted more and more attention in both academic and industrial circles. Space crowdsourcing requires workers to complete tasks at specific locations. Through crowd sourcing, requestors may post spatial tasks to a crowd sourcing server (e.g., monitoring traffic conditions or pick-up passengers) and then distribute these tasks to workers (referred to as task distribution), who complete their tasks by actually moving to a specified location. Because the spatial crowdsourcing problem is the same as many phenomenological patterns in the real world, SC is relevant to many everyday applications, including real-time taxi calling services, takeaway services, etc. In recent years, research on SC has been abundant, and many techniques for task allocation have been proposed for various application scenarios. Such as a reliable diversity-based spatial crowdsourcing (RDB-SC) problem, with the goal of maximizing the diversity of tasks. There is also a tensor decomposition based algorithm to learn the preferences of workers, assigning tasks by translating the assignment problem into a Minimum Cost Maximum Flow (MCMF) problem.
However, existing research has mainly focused on the time-space availability of workers and tasks, and thus the problem of how to efficiently and effectively assign tasks has not been substantially solved. In particular, user churn (i.e., worker churn), i.e., traitors of workers from SC service providers, is not considered in the assignment of tasks. Due to globalization and intense competition of services, the SC market tends to saturate and the cost of acquiring workers rises rapidly, and therefore, it is very critical to predict the worker's attrition rate and take some measures to retain the workers. Research on user churn begins with customer relationship management and has been proposed in various service areas. Telecommunications companies have also proposed some machine learning methods and artificial neural networks to solve this problem. In addition, banks and websites, for example, have also conducted research on user churn prediction.
Conventionally, the user churn prediction problem is regarded as a classification problem, and users are generally classified into two categories, churn and non-churn. In order to solve the problem of user churn prediction, it is necessary to take into account the characteristics of the user, including status information, behavior information, and other characteristics extracted from historical user profiles. In existing research, such as a new financial-based approach, a cost-sensitive customer churn prediction model is developed. And a two-stage loss prediction model is established to predict the loss of the user by mixing a K-means algorithm, a multi-layer perceptron artificial neural network (MLP-ANN) and a self-organizing map (SOM). However, due to the sparsity of data, spatiotemporal characteristics, and uncertain churn criteria in SC applications, the above methods cannot be directly applied to the prediction of worker churn in SC.
Disclosure of Invention
The invention aims to provide a space crowdsourcing task allocation method based on worker loss prediction. Including a worker churn prediction phase and a task allocation phase. The purpose of the first stage is to predict future worker runoff. The method comprises the steps of capturing implicit perception (LFC) of workers by using a long-short-term memory neural network (LSTM) model based on historical data, predicting idle time intervals of the workers, comparing the idle time intervals with a time threshold, marking the workers as easily lost if the time intervals exceed the time threshold, introducing a worker loss prediction model based on behaviors, converting traditional worker prediction loss serving as a classification problem into a regression problem, and defining an index to measure worker loss prediction performance. In the distribution stage, a greedy method and an optimal weight distribution (KM) -based method are provided to realize task distribution. The purpose of greedy is to greedy assign tasks to workers, where workers that are prone to run-off are given higher priority. Whereas the KM-based approach is to find the global maximum weight match on a bipartite graph (consisting of workers and tasks) taking into account the idle time intervals of workers and the reward of tasks. The invention can obviously reduce the worker loss probability in the space crowdsourcing problem and realize the highest total return of task allocation on the basis of worker loss prediction.
In order to achieve the purpose, the invention is implemented according to the following technical scheme:
the invention comprises the following steps:
s1: converting the worker loss problem into a behavior-based problem, introducing a behavior-based user loss prediction model, converting the classification problem into a regression problem by using the user loss prediction model, and predicting the idle time interval of a user;
s2: introducing a satisfaction degree related vector and a enthusiasm degree related vector into a worker loss prediction model based on implicit perception capture;
s3: the implicit perception capture is based on an LSTM model, and the implicit perception capture model is realized through two LSTMs, wherein one LSTM is used for capturing satisfaction, and the other LSTM is used for capturing enthusiasm;
s4: calculating the satisfaction degree of the current moment through the obtained satisfaction degree related vector and the previous satisfaction degree; the process of capturing enthusiasm is the same as the process of capturing satisfaction;
s5: two paths of LSTM outputs for capturing satisfaction and enthusiasm are connected in series, and a predicted idle time interval tensor is output through a full connection layer;
s6: and performing task allocation through a greedy algorithm based on worker loss or an optimal weight matching algorithm based on worker loss.
The invention has the beneficial effects that:
compared with the prior art, the invention improves the proportion of tasks allocated to the workers prone to loss by predicting the worker loss, reduces the occurrence of the worker loss in space crowdsourcing, and ensures that the total reward of the workers in task allocation is higher as much as possible. The model based on a long-short term memory neural network (LSTM) is used for extracting implicit feelings of workers, idle time intervals of the workers are learned from execution data of historical tasks of the workers, the model captures satisfaction and enthusiasm of the workers through two LSTMs, then predicted idle time interval tensors are output through a full connection layer, and the condition that the workers are prone to losing is judged according to a time interval threshold value; and then, according to a greedy algorithm sorted by idle time, adding associated edge number limitation and combining with an actual optimal weight matching (KM) task allocation algorithm of an increasing optimization strategy, the efficiency and effectiveness of task allocation in spatial crowdsourcing are improved, the experience of workers prone to loss in the spatial crowdsourcing problem is greatly improved, and therefore the occurrence of worker loss is reduced.
Drawings
FIG. 1 is a diagram of the overall architecture of worker churn prediction and task allocation according to the present invention;
FIG. 2 is a block diagram of the worker's implicit perception capture model of the present invention;
FIG. 3 is a block diagram of a satisfaction capture model based on long-short term memory neural network according to the present invention;
FIG. 4 is a graph of the impact of task number on task allocation performance;
in fig. 4: (a) CPU time; (b) distributing the total rewards obtained by the tasks; (c) the proportion of the easily lost employees of the task is distributed;
FIG. 5 is a graph of the impact of worker number on task allocation performance;
in fig. 5: (a) CPU time; (b) distributing the total rewards obtained by the tasks; (c) the proportion of the easily lost staff of the task is allocated.
Detailed Description
The invention will be further described with reference to the drawings and specific embodiments, which are illustrative of the invention and are not to be construed as limiting the invention.
As shown in fig. 1-3: the invention relates to a worker loss prediction and task allocation method based on implicit perception capture and an optimal weight matching algorithm, which is a two-stage framework and respectively comprises worker loss prediction and task allocation.
In the worker churn prediction phase, a behavior-based worker churn prediction model and a implicit perceptual capture (LFC) -based worker churn prediction model are used.
First, the problem of general user churn prediction is considered a classification problem with labeled data. However, in such fields as SC, the criteria for worker loss are not always the same, and it is difficult to judge whether a worker is actually lost. Furthermore, the marked data is not available in the SC. To solve these problems, the worker churn problem is translated into a behavior-based problem, and a behavior-based user churn prediction model (BMM-UCP) method is introduced. Different from the traditional user loss prediction method, the BMM-UCP converts the classification problem into a regression problem, predicts the idle time interval of the user, and expresses the prediction accuracy by a formula as follows:
where l (x) is an indicator function, τiIs the predicted idle time interval for worker w,
is the true value, μ is the time threshold, Λ is the logical AND operation, AND v is the logical OR operation. N is the total number of workers.
Secondly, a worker loss prediction model based on implicit perceptual capture (LFC), in practice, the idle time interval of a worker is affected by the implicit perception of the worker, such as satisfaction and enthusiasm. Therefore, the technical scheme of the invention introduces the two vectors as the input of the LFC, namely the satisfaction degree related vector and the enthusiasm degree related vector. For worker w, the satisfaction correlation vector is represented as (r)t,dt,lot,lat)，rtRepresenting the reward for the t-th task in worker w's historical data. dtRepresenting the geographical distance between the t-th task and the t-1-th task, lotRepresenting longitude, la, of the t-th tasktRepresenting the latitude of the t-th task. Enthusiasm is a feeling similar to satisfaction with (r)t,wt,lot,lat) And (4) showing. Wherein, wtRepresenting the time interval between the t-th task and the t-1 th task in the historical data of the worker w, and the meanings of the other three elements are the same as the corresponding elements in the satisfaction degree correlation vector. LFC is a model based on LSTM, so the present invention sets the time step to 10, i.e., one input instance consists of 10 satisfaction-related vectors and 10 enthusiasm-related vectors.
The idle time interval is related to the implicit feeling of the worker. It is apparent that the satisfaction and enthusiasm of a worker after performing the tth task is related to his satisfaction and enthusiasm before performing the tth task and some attributes of the tth task. The invention considers the implicit feelings (such as satisfaction and enthusiasm) as sequential data, which means that the implicit feelings have sequential dependency. The LFC model is realized through two paths of LSTMs, wherein one path is used for capturing the satisfaction degree, and the other path is used for capturing the enthusiasm degree.
For historical data of worker w, use Sw＝{s1,s2,…,snDenotes that the task s is executedtThe degree of satisfaction of the latter depends on the spatial task stAnd satisfaction before execution. Therefore, the invention can calculate the satisfaction degree of the current moment by the obtained satisfaction degree related vector and the previous satisfaction degree:
where NL represents a non-linearity,
i.e. the satisfaction of the worker after completing the t-th task.
The calculation process of enthusiasm is the same as that of satisfaction:
wherein,
is the enthusiasm of the worker after the task of the t-th item is completed.
As mentioned above, satisfaction and enthusiasm are both sequentially dependent. LSTM solves the long term dependency problem with three gate mechanisms and performs well in processing sequential data. Therefore, the invention uses the LSTM-based model for satisfaction capture, as shown in the following formula:
wherein,
represents the content to be forgotten and is calculated using sigmoid as an activation function,
a vector relating to the degree of satisfaction is represented,
indicating the satisfaction of the last time step after filtering.
Representing content to be updated, its input and activation functions and calculations
The same, but different parameters.
Is a candidate satisfaction degree, which is
And satisfaction after filtration
As input, activation is performed with a tanh function,
representing the final output of the model. The process of capturing enthusiasm is similar to the process of capturing satisfaction.
And finally, connecting two paths of LSTM outputs for capturing satisfaction and enthusiasm in series, and outputting the predicted idle time interval tensor X through a full connection layer. The expression formula is as follows:
X＝[τ1,τ2,τ3,…]
wherein tau iswIs the predicted idle time interval of worker W, WspIs a weight matrix, bspIs a deviation.
By giving a time threshold μ, if the tensor X [ w ] corresponding to worker w exceeds μ, worker w is determined to be a worker who is vulnerable to loss.
In the task allocation stage, two task allocation algorithms are provided, namely a greedy algorithm based on worker loss and an optimal weight matching (KM) algorithm based on worker loss.
And on the basis of the idle time interval of each worker obtained by LFC model output, sequencing the available workers of each task in a descending order by a greedy algorithm based on worker loss, and distributing the task to the worker with the largest idle time interval. There are two constraints on the available workers: firstly, the distance between a worker and a task does not exceed the reachable distance of the worker; the second is that the current time plus the time required for the worker to complete the task does not exceed the expiration time of the task.
And converting the task allocation problem into a bipartite graph maximum weight matching problem by establishing a bipartite graph of the workers and the tasks based on the KM algorithm of worker loss. The point set in the created bipartite graph is divided into VWAnd VSTwo sets, each worker wiMapping to a vertex
Each space task sjMapping to a vertex
The edge sets are added according to space-time limits, and workers are driven
Connecting to space tasks
The weight of the edge is given by the time interval tauiAnd spatial tasks sjThe awards are respectively obtained by weighted calculation, and the formula is as follows:
wherein wcWeight representing time interval, wrWeight, s, representing reward for a spatial taskjR denotes a spatial task sjThe prize of (1).
Then each worker is defined
Is the adaptive upper limit u of the number of associated edgesiSimilarly, each task also defines an upper bound on its associated number of edges. Wherein u isi＝ρ*X[i]And rho is a hyperparameter. The purpose is to improve the efficiency of task allocation.
To better implement the KM-based task assignment algorithm, the invention first implements a function that recursively finds a worker for a task that can be matched, calculates the difference between the weight of the edge associated with the two vertices and the sum of the expected values of the worker and the task, and if the difference is equal to 0, the task can be assigned to the worker. Where the expected values for workers and tasks are the maximum weight values in the edges associated therewith. When the workers fail to match any task, the expectations of the workers and tasks involved in the last matching are adjusted to change the competition among the workers, so that the aim of distributing more workers is fulfilled.
Furthermore, since the worker task bipartite graph may not have the reality of a perfect match, for workers with an expected value less than 0, the invention will stop matching tasks for that worker.
The invention uses check-in data set from Yelp to carry out experiment, and in order to make user data more representative, the invention filters the data and selects the user data with the number of comments exceeding 20 and the number of comments exceeding 10 before 2019-05-1423:22: 59.
In the aspect of worker loss prediction, the prediction accuracy of the method is compared with that of a Linear Regression (LR) method and a multilayer fully-connected neural network (MC) method under different time thresholds, and the result is shown in a table 4-1. The superiority of implicit perceptual capture (LFC) in predicting worker loss can be demonstrated.
TABLE 4-1 accuracy of worker loss prediction
In the aspect of task allocation, the invention compares the performances of four algorithms under three different indexes, wherein the four algorithms are as follows: greedy task allocation algorithm (Greedy) that does not consider worker churn; matching task allocation algorithm (KM) with optimal weight without considering worker loss; a Greedy task allocation algorithm (Greedy + WC) for predicting worker loss based on implicit sensing capture; and capturing and predicting the optimal weight matching task allocation algorithm (KM + WC) of worker loss based on implicit perception. The three indexes are respectively: finding a CPU time cost (CPU time) of the task allocation; total rewarded for the assignment of tasks; a proportion of churning employees (Assignment ratio) of the task is assigned. The experimental results are shown in fig. 4 and 5. Wherein, fig. 4(a) shows the positive correlation between the number of tasks and the three indexes, and fig. 4(b) shows that the number of workers is positively correlated with the first two indexes and negatively correlated with the proportion of the easily lost workers allocated to the tasks. It can be shown that the method considering worker loss is superior to that not considered, and the higher proportion of the easily lost workers allocated with tasks indicates that the tasks allocated to the easily lost workers are more, so that worker loss can be effectively reduced, and meanwhile, the total reward obtained by task allocation is higher, so that the benefit is maximized.
The experimental result shows that the method used by the invention has better performance than the common algorithm, and particularly has obvious improvement on the distribution ratio of the total reward to the employees easy to lose.
The technical solution of the present invention is not limited to the limitations of the above specific embodiments, and all technical modifications made according to the technical solution of the present invention fall within the protection scope of the present invention.
Claims (9)
1. A spatial crowdsourcing task allocation method based on worker loss prediction is characterized by comprising the following steps:
s1: converting the worker loss problem into a behavior-based problem, introducing a behavior-based user loss prediction model, converting the classification problem into a regression problem by using the user loss prediction model, and predicting the idle time interval of a user;
s2: introducing a satisfaction degree related vector and a enthusiasm degree related vector into a worker loss prediction model based on implicit perception capture;
s3: the implicit perception capture is based on an LSTM model, and the implicit perception capture model is realized through two LSTMs, wherein one LSTM is used for capturing satisfaction, and the other LSTM is used for capturing enthusiasm;
s4: calculating the satisfaction degree of the current moment through the obtained satisfaction degree related vector and the previous satisfaction degree; the process of capturing enthusiasm is the same as the process of capturing satisfaction;
s5: two paths of LSTM outputs for capturing satisfaction and enthusiasm are connected in series, and a predicted idle time interval tensor is output through a full connection layer;
s6: and performing task allocation through a greedy algorithm based on worker loss or an optimal weight matching algorithm based on worker loss.
2. The worker churn prediction based spatial crowd-sourced task allocation method according to claim 1, wherein: the accuracy of predicting the idle time interval of the user in step S1 is formulated as follows:
where l (x) is an indicator function, τiIs the predicted idle time interval for worker w,
is the true value, μ is the time threshold, Λ is the logical AND operation, V is the logical OR operation; n is the total number of workers.
3. The worker churn prediction based spatial crowd-sourced task allocation method according to claim 1, wherein: in the step S2, in the above step,for worker w, the satisfaction correlation vector is represented as (r)t，dt，lot，lat)，rtA reward representing the t-th task in the historical data of worker w; dtRepresenting the geographical distance between the t-th task and the t-1-th task, lotRepresenting longitude, la, of the t-th tasktRepresenting the latitude of the t-th task; enthusiasm is a feeling similar to satisfaction with (r)t，wt，lot，lat) Represents; wherein, wtRepresenting the time interval between the t-th task and the t-1 th task in the historical data of the worker w, and the meanings of the other three elements are the same as the corresponding elements in the satisfaction degree correlation vector.
4. The worker churn prediction based spatial crowd-sourced task allocation method according to claim 1, wherein: in the step S3: for historical data of worker w, use Sw＝{s1，s2，...，snDenotes that the task s is executedtThe degree of satisfaction of the latter depends on the spatial task stAnd satisfaction before execution.
5. The worker churn prediction based spatial crowd-sourced task allocation method according to claim 1, wherein: in step S4, the satisfaction of calculating the current time is:
where NL represents a non-linearity,
the satisfaction degree of the worker after the t task is completed is obtained;
the calculation process of enthusiasm is the same as that of satisfaction:
wherein,
the enthusiasm of workers after the t-th task is completed;
satisfaction capture was performed using an LSTM-based model, as shown in the following equation:
wherein,
represents the content to be forgotten and is calculated using sigmoid as an activation function,
is shown andthe vectors that are related to the degree of satisfaction,
representing the satisfaction after the last time step filtering;
representing content to be updated, its input and activation functions and calculations
The same time, but different parameters;
is a candidate satisfaction degree, which is
And satisfaction after filtration
As input, activation is performed with a tanh function,
representing the final output of the model.
6. The worker churn prediction based spatial crowd-sourced task allocation method according to claim 1, wherein: in step S5, two LSTM outputs of the capturing satisfaction and the capturing enthusiasm are connected in series, and a predicted idle time interval tensor X is output through a full connection layer; the expression formula is as follows:
X＝[τ1，τ2，τ3，...]
wherein tau iswIs a prediction of worker wIdle time interval, WspIs a weight matrix, bspIs a deviation;
by giving a time threshold μ, if the tensor X [ w ] corresponding to worker w exceeds μ, worker w is determined to be a worker who is vulnerable to loss.
7. The worker churn prediction based spatial crowd-sourced task allocation method according to claim 1, wherein: in the step S6, based on the worker-loss greedy algorithm, on the basis of the idle time interval of each worker obtained by the implicit sensing capture model, sorting the available workers of each task in a descending order, and allocating the task to the worker with the largest idle time interval; there are two constraints on the available workers: firstly, the distance between a worker and a task does not exceed the reachable distance of the worker; the second is that the current time plus the time required for the worker to complete the task does not exceed the expiration time of the task.
8. The worker churn prediction based spatial crowd-sourced task allocation method according to claim 1, wherein: in the step S6, the task allocation problem is converted into a bipartite graph maximum weight matching problem by establishing a bipartite graph of workers and tasks based on an optimal weight matching algorithm for worker loss; the point set in the created bipartite graph is divided into VWAnd VSTwo sets, each worker wiMapping to a vertex
Each space task sjMapping to a vertex
The edge sets are added according to space-time limits, and workers are driven
Connecting to space tasks
The weight of the edge is given by the time interval tauiAnd spatial tasks sjThe awards are respectively obtained by weighted calculation, and the formula is as follows:
wherein wcWeight representing time interval, wrWeight, s, representing reward for a spatial taskjR denotes a spatial task sjThe reward of (1);
then defining each worker
Is the adaptive upper limit u of the number of associated edgesiSimilarly, each task also defines an upper limit of its associated number of edges; wherein u isi＝ρ*X[i]Rho is a hyper-parameter; the purpose is to improve the efficiency of task allocation.
9. The worker churn prediction based spatial crowd-sourced task allocation method of claim 8, wherein: in the optimal weight matching algorithm for worker loss, a function for recursively searching for a task capable of being matched for a worker is realized, the function calculates a difference value between the weight of an edge associated with two vertexes and the sum of expected values of the worker and the task, and if the difference value is equal to 0, the task can be allocated to the worker; wherein the expected values for the workers and tasks are the maximum weight values in the edges associated therewith; when the worker fails to match any task, the expectation of the worker and the task involved in the last matching is adjusted to change the competition relationship among the workers, so that the aim of distributing more workers is fulfilled; for a worker whose expected value is less than 0, the task matching for that worker is stopped.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111589692.4A CN114372680A (en) | 2021-12-23 | 2021-12-23 | Spatial crowdsourcing task allocation method based on worker loss prediction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111589692.4A CN114372680A (en) | 2021-12-23 | 2021-12-23 | Spatial crowdsourcing task allocation method based on worker loss prediction |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114372680A true CN114372680A (en) | 2022-04-19 |
Family
ID=81142043
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111589692.4A Pending CN114372680A (en) | 2021-12-23 | 2021-12-23 | Spatial crowdsourcing task allocation method based on worker loss prediction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114372680A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115694877A (en) * | 2022-08-30 | 2023-02-03 | 电子科技大学长三角研究院(衢州) | Space crowdsourcing task allocation method based on federal preference learning |
| CN117273324A (en) * | 2023-09-15 | 2023-12-22 | 国网湖北省电力有限公司黄石供电公司 | Staff task allocation excitation algorithm based on personalized long-short time model |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103295148A (en) * | 2012-02-27 | 2013-09-11 | 埃森哲环球服务有限公司 | Digital consumer data model and customer analytic record |
| US20140165071A1 (en) * | 2012-12-06 | 2014-06-12 | Xerox Corporation | Method and system for managing allocation of tasks to be crowdsourced |
| CN109033865A (en) * | 2018-06-20 | 2018-12-18 | 苏州大学 | The method for allocating tasks of secret protection in a kind of space crowdsourcing |
| CN110400128A (en) * | 2019-07-29 | 2019-11-01 | 电子科技大学 | A kind of space crowdsourcing method for allocating tasks based on the perception of worker's preference |
| CN111311115A (en) * | 2020-03-12 | 2020-06-19 | 电子科技大学 | Group task allocation method based on space crowdsourcing social influence preference |
| CN112328914A (en) * | 2020-11-06 | 2021-02-05 | 辽宁工程技术大学 | Task allocation method based on space-time crowdsourcing worker behavior prediction |
| CN113627765A (en) * | 2021-08-01 | 2021-11-09 | 湖南大学 | User satisfaction-based distributed space crowdsourcing task allocation method and system |
-
2021
- 2021-12-23 CN CN202111589692.4A patent/CN114372680A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103295148A (en) * | 2012-02-27 | 2013-09-11 | 埃森哲环球服务有限公司 | Digital consumer data model and customer analytic record |
| US20140165071A1 (en) * | 2012-12-06 | 2014-06-12 | Xerox Corporation | Method and system for managing allocation of tasks to be crowdsourced |
| CN109033865A (en) * | 2018-06-20 | 2018-12-18 | 苏州大学 | The method for allocating tasks of secret protection in a kind of space crowdsourcing |
| CN110400128A (en) * | 2019-07-29 | 2019-11-01 | 电子科技大学 | A kind of space crowdsourcing method for allocating tasks based on the perception of worker's preference |
| CN111311115A (en) * | 2020-03-12 | 2020-06-19 | 电子科技大学 | Group task allocation method based on space crowdsourcing social influence preference |
| CN112328914A (en) * | 2020-11-06 | 2021-02-05 | 辽宁工程技术大学 | Task allocation method based on space-time crowdsourcing worker behavior prediction |
| CN113627765A (en) * | 2021-08-01 | 2021-11-09 | 湖南大学 | User satisfaction-based distributed space crowdsourcing task allocation method and system |
Non-Patent Citations (4)
| Title |
|---|
| YAN ZHAO ET AL.: "Predictive Task Assignment in Spatial Crowdsourcing: A Data-driven Approach", 《2020 IEEE 36TH INTERNATIONAL CONFERENCE ON DATA ENGINEERING (ICDE)》 * |
| ZIWEI WANG ET AL.: "Task Assignment withWorker Churn Prediction in Spatial Crowdsourcing", 《 IN PROCEEDINGS OF THE 30TH ACM INTERNATIONAL CONFERENCE ON INFORMATION AND KNOWLEDGE MANAGEMENT (CIKM ’21)》 * |
| 李洋等: "基于树分解的空间众包最优任务分配算法", 《软件学报》 * |
| 林荟荟等: "空间众包中可拒绝情况下的在线任务分配", 《浙江科技学院学报》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115694877A (en) * | 2022-08-30 | 2023-02-03 | 电子科技大学长三角研究院(衢州) | Space crowdsourcing task allocation method based on federal preference learning |
| CN115694877B (en) * | 2022-08-30 | 2023-08-15 | 电子科技大学长三角研究院(衢州) | Space crowdsourcing task allocation method based on federal preference learning |
| CN117273324A (en) * | 2023-09-15 | 2023-12-22 | 国网湖北省电力有限公司黄石供电公司 | Staff task allocation excitation algorithm based on personalized long-short time model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109191896B (en) | Personalized parking space recommendation method and system | |
| CN103473291B (en) | Personalized service recommendation system and method based on latent semantic probability models | |
| Bobbio et al. | Analysis of large scale interacting systems by mean field method | |
| CN113313947A (en) | Road condition evaluation method of short-term traffic prediction graph convolution network | |
| CN111611488B (en) | Information recommendation method and device based on artificial intelligence and electronic equipment | |
| CN114372680A (en) | Spatial crowdsourcing task allocation method based on worker loss prediction | |
| CN113298191B (en) | User behavior identification method based on personalized semi-supervised online federal learning | |
| CN113762595A (en) | Traffic time prediction model training method, traffic time prediction method and equipment | |
| Li et al. | Research on QoS service composition based on coevolutionary genetic algorithm | |
| Qian et al. | Short-term demand forecasting for on-demand mobility service | |
| Ye et al. | Bridging the micro and macro: Calibration of agent-based model using mean-field dynamics | |
| Zhao et al. | Unifying Uber and taxi data via deep models for taxi passenger demand prediction | |
| Handoyo et al. | The fuzzy inference system with rule bases generated by using the fuzzy C-means to predict regional minimum wage in Indonesia | |
| CN109657725B (en) | Service quality prediction method and system based on complex space-time context awareness | |
| Shinkuma et al. | Data assessment and prioritization in mobile networks for real-time prediction of spatial information using machine learning | |
| Sefati et al. | Meet User’s Service Requirements in Smart Cities Using Recurrent Neural Networks and Optimization Algorithm | |
| CN113516229A (en) | Credible user optimization selection method facing crowd sensing system | |
| Zhao et al. | Worker-churn-based task assignment with context-lstm in spatial crowdsourcing | |
| KR20210042709A (en) | Method and server for stock movement prediction using corporate relation data | |
| CN114462707A (en) | Web service multidimensional QoS (quality of service) joint prediction method based on feature depth fusion | |
| Chu et al. | Deep Encoder Cross Network for Estimated Time of Arrival | |
| Nguyen et al. | Data-driven models for individual and group decision making | |
| Divya et al. | DRGNN-DILATED RECURRENT GRAPH NEURAL NETWORK FRAMEWORK INCORPORATING SPATIAL AND TEMPORAL FEATURES SIGNIFYING SOCIAL RELATIONSHIPS IN IOT NETWORK BASED TRAFFIC PREDICTION. | |
| CN113450557B (en) | Method and device for updating prediction model for passenger flow of vehicle | |
| CN116050640B (en) | Short-time passenger flow prediction method of multi-mode traffic system based on self-adaptive multi-graph convolution |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20220419 |
|
| RJ01 | Rejection of invention patent application after publication |