Application diversity and good user perception are key to LTE network operation as well as the future 5 G network operation. The current KPI index system is mainly used to assess the status of network

which can not refl ects the actual level of user’s satisfaction on networks and applications. One method of user perception evaluation of subdivision applications is proposed. Firstly

the KQI indicators of LTE applications and the relevant data of XDR

MR are analyzed to extract the QoE evaluation features of specific business. Then

by using machine learning

the customer perception evaluation model of differentiated applications is created. Finally

as an example

the process of selecting QoE evaluation features and creating the model of video business are illustrated. This method can precisely evaluate user perception in finer-grained. Therefore

it provides accurate data support to network optimization and network operation based on customer perception.