The invention relates to an optimization algorithm based on edge computing that is applied in the MCS (Mobile Crowd Sensing) sensing task scene, and is based on the recruitment of task participants, the judgment of information validity, and the detection and correction of errors. The present invention represents platform utility as a logarithmic function (strictly increasing and sagging) of independent measurement data for each location, and the algorithm aims to maximize total utility under exerting budget constraints. Since the proposed joint optimization problem is a non-deterministic polynomial combinatorial optimization problem, it cannot be solved in polynomial time. Therefore, we design an approximation algorithm that greatly reduces the computational complexity. The present invention uses a deep neural network (DNN) for data verification, thereby increasing the accuracy of predicted data. In addition, the present invention also adopts data redundancy to improve the sensing quality, and when measurement errors are detected near the data source, the amount of data transmitted from the base station to the sensing platform is greatly reduced, the occupied frequency band resource is reduced, and the time delay is reduced.