Research On The Algorithm Of Improving User Network Satisfaction

As China's telecommunication market gradually enters the mature stage,the focus of market competition has shifted from the incremental market to the stock market.The operators have begun to actively change user retention strategies and strengthen the user-centric development concept.Therefore,the research on net-work satisfaction improving algorithm around user perception is of great sig-nificance in the intelligent transformation of operators.In response to user complaints,the traditional method of taking post-mortem remedies seriously affects user perception.Network complaints are not only related to user experience,but also to network maintenance,which is particularly important in operators' complaint management.To this end,this article focuses on the early warning and response of network complaints,uses artificial intelligence to locate potential complaining users and formulate care priorities,and finally improve user satisfaction by reducing complaints.The main work and innovations made in this article are as follows:Different user groups have different characteristics and demands.Aiming at this problem,this paper studies the problem of user clustering.This paper proposes a metric function to evaluate the effect of clustering,and also proposes a user clustering algo-rithm based on semi-supervised clustering,which combines expert experience with data mining algorithms,and iterates and divides based on density peaks under existing constraints.Experiments show that the proposed algorithm has low time complexity,the clustering result is more balanced and effective,and it matches the operator's actual work more.Based on the result of user clustering,the concept of sensitivity factor is proposed for feature con-struction.Aiming at the research of network complaint warning,this paper improves the Stacking integrated classification algorithm,proposes a T×T fold-ing loop training method,avoids cross-learning problems,and combines the advantages of multiple base classifiers based on structural features and original features to build a network complaint warning model.The net loss function is proposed to assist the decision of the optimal decision threshold.Experiments show that the proposed algorithm can effectively improve prediction performance,locate potential complaining users more accurately,and reduce the losses of the enterprise.Aiming at the research on the priority of potential complaining users,this paper combines network performance indicators,user service usage needs and user scoring data to build a network performance scorecard model based on user perception,to quantitatively predict user network satisfaction and finally outputs a corresponding satisfaction score priority.Experiments show that the model can predict high-risk complaining users in a specified user group based on satisfaction scores,which provides a theoretical basis for operators to build a hierarchical user care system.This paper provides theoretical support for reversing the current complaint handling method,and also provides ideas for improving user network satisfaction.