Research On Automatic Assignment Of Crowdsourcing Platform Task Based On The Genetic Nonnegative Matrix Decomposition Algorithm

The crowdsourcing is a new way of solving problems after outsourcing,which refers to a way in which an organization assigns work which should be done by the employee to a non-specific mass on internet and they accept the work voluntarily.With the crowdsourcing model used widely,the task on the crowdsourcing platform has exploded,however,the simple task scheduling method of the crowdsourcing platform does not meet the growing demand for tasks,resulting in the task of the platform can not be completed in a timely manner.Therefore,the task allocation algorithm of the crowdsourcing platform is proposed,and the appropriate tasks are automatically assigned to the appropriate users,which will greatly improve the task completion efficiency of the platform.In recent years,domestic and foreign researchers have proposed the use of matrix decomposition algorithm to carry out automatic assignment of tasks.Especially non-negative matrix decomposition has gained tons of attention because of its advantages,such as good explanation,accuracy and alleviating the cold start problem.However,the objective function of the nonnegative matrix decomposition algorithm is usually non-differentiable and discontinuous,and the gradient search method is easy to fall into the local optimal.Based on this,this paper proposes a non-negative matrix decomposition algorithm based on genetic algorithm to realize the automatic allocation of the task in crowdcourcing platform,and the global optimality of the genetic algorithm is used to improve the accuracy of the algorithm.In the initial stage of matrix decomposition,user characteristic matrix and task characteristic matrix are initialized by using the Frobenius norm of the row vector and the column vector of the difference of the original matrix and the approximation matrix respectively.On the basis of this,the crossover of the matrix random rows or columns and matrix fixed proportionality element variation are used to minimize the Frobenius norm of the difference of theoriginal matrix and the approximate matrix that is initialized.After the number of fixed iterations of the two nonnegative matrices,the product of the two nonnegative matrices is the approximate matrix of the original matrix.Given a specified number of tasks,we predict the missing values of the original matrix to sort the task,assigning tasks with high predictive ratings to the user.In this paper,the non-negative matrix decomposition algorithm based on genetic algorithm is compared with classical probability matrix decomposition algorithm,random initialized NMF algorithm and TaskRec algorithm for RMSE and MAE precision analysis.The simulation results show that the RMSE and MAE indexes better,with a higher degree of accuracy.