Research On Key Technology Of IPTV User Experience Improvement Based On User Interest Mining

With the development of broadband technology,the dissemination of multimedia content through the network has become more and more efficient.A typical application scenario brought about by this is interactive Internet Protocol Television(IPTV).In recent years,IPTV has become more and more popular.According to the latest data from the Ministry of Industry and Information Technology,to the end of June 2018,the number of IPTV users developed by basic telecommunications companies reached 142 million,a net increase of 2,200,000 in the first half of the year,and the user scale continued to grow steadily.As a result,various types of services in the future IPTV network emphasize not only these network indicators such as transmission speed and bandwidth,but also the overall experience of users in accepting these services.In this thesis,we mainly study the key technology improvement methods of quality of user experience based on user interest mining.The main research work of the thesis is mainly reflected in the following three aspects:Firstly,based on statistical knowledge,the data preprocessing work is proposed,and a series of data preprocessing methods are proposed,including data cleaning,outlier identification and processing,and data normalization.Then,when analyzing the behavior of IPTV users watching TV,two modeling methods for user interest mining are proposed.The first one is based on the label of TV program and the viewing time of users,and proposes a Gaussian Mixture Model.The clustering algorithm referred to as GMM generates the characteristics of the user's viewing behavior sequence to represent the user's interest pattern of watching TV programs,which is convenient for analysis and further use of the feature;the second is the recurrent neural network based on the attention mechanism which is used as user interest mining model.This model directly generates the user's interest score in the current TV program according to the user's historical viewing record.Experimental results show that both modeling methods can effectively mine user features.Finally,an improved artificial neural network prediction model is proposed to predict the quality of user experience.Firstly,the key factors affecting the user experience are analyzed,which are mainly divided into two aspects.One is the influencing factors at the service level,mainly refers to the content,network and equipment related factors,and the other is the user-level influencing factors such as user interest.Then based on the user's viewing behavior sequence characteristics and user interest characteristics,two neural network structures are proposed respectively.Finally,the proposed model is compared with the traditional support vector machine and decision tree model.The experimental results show that the user experience quality is closely related to user interest and the proposed improved neural network user experience prediction model has higher accuracy.