The invention discloses a network perception intelligent early warning and improvement method based on big data, specifically: first, classify the source data, and analyze the five aspects of network quality, business perception, complaint traceability, terminal adaptation, and user behavior according to the degree of influence. The user perception index sets the scoring ratio to obtain the online satisfaction score, and then the user's online data goes through seven stages of attach, pdn, tau, dns analysis, tcp three-way handshake link establishment, HTTP request / response, sp response / transmission, Find out the reasons that affect the low score of online satisfaction, and judge that it belongs to the specific process in the seven-stage process; finally, through segmentation and demarcation, cluster the problems into the wireless side, core network side, terminal side, and content side, and target Issues are dealt with accordingly. This method has comprehensive data analysis, can obtain the real perception of the user, improves the correlation between the quality of the network indicators and the real perception of the user, and improves the customer satisfaction of 4G Internet access.