Providing good user perception is the objective of telecom operators' mobile network operation and optimization. Specifically, telecom operators try to improve the quality of mobile network to meet telecom customers' business needs. However, telecom operators envisage a key problem that the statistics from the network dimension do not directly reflect the user perception of the network. This results in the disconnection between the network operation and the actual user perception. This paper mainly discusses the relationship between key network indicators and user perception. Then, this paper realizes the prediction of user perception from the network side through random forest algorithm. The proposed user perception algorithm can effectively assist the network operation and maintenance personnel to optimize the network.