Research On Network User Preference Analysis And Topic Evolution Trend Forecast

The rapid development of network technology makes the impact of the Internet on social life growing. Internet users as the main information dissemination on the Internet, has a direct impact on their behavior patterns process for the dissemination of information on the Internet. The user preferences for the topic is one of factors which can affect the trend of the topic, it can contribute to the technology of Internet topic trend forecast. The explosive growth of the Internet information increases the difficulty of Internet users on the Internet to obtain information. To improve the efficiency of Internet users in the Internet to retrieve information, it needs to be analyzed on information index and users preferences for information, and to provide personalized information services according to user preference. This paper integrates the interdisciplinary research methods and ideas, starting from the analysis of user behavior patterns, to analyze the preferences of Internet users, analysis of Internet information dissemination process, and in the end, it establishes an effective Internet topic trend forecast model. It starts from the point of the user preferences, and researches the information recommendation mechanism to provide users with a more convenient and efficient information retrieval. The work of the dissertation is supported by the National Natural Science Foundation of China (No.61172072), National Natural Science Foundation of China (No.61271308), Beijing Natural Science Foundation (No.4112045), Specialized Research Fund for the Doctoral Program of Higher Education of China (No.20100009110002) and the Academic Discipline and Postgraduate Education Project of Beijing Municipal Commission of Education.Major works and innovations of the paper include the following aspects:1) Take micro blog, one typical use of the Internet, for example, analyze the statistical properties of the user’s behavior, and lay the foundation for analysis of user preference and topic evolution trend. Concerned about the proportion of the ability of the user to the dissemination of information by the number of user followers (dissemination width), comment and repost number of user status (importance of information), and the percent of followers of comments of user status (dissemination depth). With the growing number of status, the proportion of new words included in the newly status is declining. The words used by the user are included in a relatively stable habits word collection. The collection of same words between friend and user of habits words collection is much larger than randomly selected user, using the customary words in the intersection of the collection quantity may be substantially the performance of the similarity between two users. Topic generated time series has significant relationship with1lag period time series, and using the correlation between the before and after of the time series can forecast the development trend of the topic.2) Take Micro blog for example, and analysis of user preferences for the content. The user’s preferences is one of the factors of the topic evolution trends, and it is also one of the factors of Recommended system, analysis of the recommendation system can contribute to the two questions above. User preference for the content can be shown by the similarity of between the contents. This paper analysis the similarity between users, topics and statuses, and these similarities are used to indicate the user preference on the other users, topic and status. The similarity between the content is from the words that contains in the content. This paper defines the decision process of user preference for the new user, topics and statuses to determine whether a user has preference for these objects, and the results can be used for the recommended model and the topic trend forecasting model.3) This paper presents an improved recommendation algorithm based on the object’s inherent attribute similarity, combined with the user’s preference algorithm applied to the micro blog system recommendation. Improved Algorithm is based on collaborative filtering algorithm, and it uses of the similarity of the intrinsic properties of the object to improve the initial rating and user similarity calculation process, so different object has a different initial rating, the similarity between two users for different object is not the same, too. The improved algorithm figures out the problem of low accuracy of data sparsity, and it has improved accuracy compared to classical algorithm. Micro blog Recommended system combines the user’s preference, the similarity between the topics, the similarity between the statuses, and it has good recommendations effect.4) This paper presents an Internet topic trend forecast algorithm. The algorithm is based on the correlation between the time series, drawing on the economics of the ARIMA model, and it can forecast topic trends in different types of media on the Internet. Analysis of the factors that affect the accuracy of forecast in the micro blog, the results of the analysis pointed out that for the statues posted by preference users, the probability of that its comments are preference statuses, is much larger than the state of the non-preference user. According to this feature, a new algorithm for the generation of the time series is presented, according to the user’s preferences on topics given a different weight value to characterize the affect of comments the for Discussions development. Use the ARIMA model for the new time series to forecast, and the forecast results show that, using the new time series model prediction error is less than using the model of the original sequence. The new series has a better effect to forecast the trend of the topic, and it is more suitable to topic trend forecast.