Task Recommendation Based On The Influencing Factors Of Crowdsourcing Participating Willingness

With the growing number of workers and requesters in crowdsourcing contest platforms,the problem of information overload becomes increasingly serious.How to effectively recommend tasks for workers has become an urgent issue to be solved.The traditional task recommendation method constructs the model of workers'interest only take the workers'bidding records of tasks into consideration,and recommend tasks to workers based on this method.However,whether the workers bid for tasks or not is essentially determined by the workers'participation willingness.Therefore,based on the study of traditional recommendation methods and the factors influencing workers'participation willingness in crowdsourcing contest,this thesis proposes a novel task recommendation method for crowdsourcing contest,which takes factors influencing workers'participation willingness into consideration,and supplements the corresponding method to alleviate the problem of data sparseness in task recommendation process.First,based on the research to factors influencing workers'participation willingness and the user-based collaborative filtering framework,the influencing factors are represented as income preference,quality and ability of workers,and the trust in requestors,from the demission of workers and requestor respectively.The influencing factors are measured on the basis of workers'historical records and related description information,as well as the worker model.Secondly,according to the worker model,the adjustable coefficient?_i??_i is introduced to calculate the similarity of the workers in the expected benefits,quality and ability,and the trust of the requestor.At the same time,the fusion coefficient?_iis introduced to calculate the comprehensive similarity of the workers.On this basis,the k-neighbor set of the workers is determined and the task recommendation list is generated according to the weighted average method.At the same time,in order to alleviate data sparseness existing in the recommendation process of the above task recommendation methods,this study proposes a neural network task recommendation method that integrates k-nearest neighbor.In this method,the worker feature and task feature are firstly processed through the feature embedding layer,then the worker vector and task vector in the worker's k-nearest neighbor are taken as the input of the neural network input layer,and the model training is carried out through the historical bidding records of workers in the training data set.Finally,a task recommendation list is generated according to the trained model under the condition of data sparseness.Based on the real data of epwk.com,numerical experiments were carried out to verify the effectiveness of our proposed task recommendation for workers and related worker model which comprehensively considering the factors influencing the workers'partition willingness,such as expected earnings,quality and ability,and trust in the employer.At the same time,the neural network task recommendation method integrating worker k-nearest neighbor can effectively alleviate the problem of data sparseness in the task recommendation process.In the construction of task recommendation model,it is necessary to process the worker and task features through the feature embedding layer.In this thesis,the influencing factors of workers'participation willingness are considered into task recommendation,which expand the traditional recommendation methods applied in the field of crowdsourcing and enrich the research of crowdsourcing task recommendation.At the same time,it provides an effective solution to solve the problem of crowdsourcing contest task recommendation,and improve the trading efficiency of knowledge and labor in crowdsourcing.