Abstract: In the user perceived feedback calculation process, the traditional feedback algorithm in multi-level user perception inconsistent cases, the difference of data will make sense the result classification swing, leading to perception feedback results are not allowed.In order to solve this problem, this paper puts forward a data set based on differentiation layered fusion user perception feedback algorithm, through the QOE-KQI-KPI three layer system model for users' perception feedback results and classification, calculation multilayer system perception feedback clustering weighted average, set each level of feedback weighted average, eliminate multiple data differences, solve the data difference too big to bring the perception of the problem of calculation is not allowed.Experiments show that the method can effectively fusion multiple users' perception results, and satisfactory results were obtained.