Personalized Crowdsourcing Task Recommendation Mechanism Based On Reliability

With the popularization of Internet technology and the development of crowdsourcing ideas,more and more crowdsourcing applications based on the Internet have entered the public's vision.In the crowdsourcing platform,the platform publishes tasks in a timely manner,and registered users actively receive tasks and report data as required.Finally,platform obtains the result of each task based on the data submitted by users.In such a system,how to improve the user's participation experience to ensure sufficient data and improve the final service quality of the task has always been the core problem in the crowdsourcing system.This paper proposes a personalized crowdsourcing task recommendation mechanism based on users' reliability.The mechanism includes two parts: recommendation algorithm and truth inference algorithm.In the recommendation algorithm part,the users' preference and the reliability are considered.User's preference score and reliability score are calculated respectively,which will be combined to get the user's recommendation score.In the truth inference part,in order to improve the final completion quality of the crowdsourcing task,this paper proposes a truth inference method that initializes the initial weights with reliability,which reduces the initialization time of the truth inference process,and provides a more accurate reliability reference for the recommended algorithm part.By combining the recommendation algorithm and the truth inference algorithm,the mechanism proposed in this paper not only guarantees the enthusiasm of users for the crowdsourcing system,but also ensure the quality of tasks.At the same time,considering the problem of the fact that the type of crowdsourced application tasks is fuzzy,the collaborative filtering algorithm cannot be applied and the amount of data is huge.This paper also adds a singular value decomposition method based on the above mechanism.The matrix is decomposed to obtain user and task vectors to calculate the predicted recommendation score and credibility score,which can be effectively applied to the actual crowdsourcing data set.Finally,through theoretical analysis and simulation experiments,we compare the proposed algorithm with some existing methods,and finally proves the efficiency,rationality and feasibility of the mechanism proposed in this paper.