Microblogging User Oriented Potential Interest Analysis

The rapid development of Internet makes network applications emerging.Microblogging as a convenient social networking platform gets people’s favoritewhile it is launched, and becomes a main platform to record people’s life, expressemotions, obtain information, make friends at work and in life. However, with therapidly growing popularity of microblogging, the amount of information inmicroblogging is also increasing quickly. A huge amount of information makespeople have to spend a lot of time to find the information they are interested.The main research topic is: how obtain microblogging users’ potential interestfrom the tweets of microblogging users. The potential interest can be used toprovide personalized services to users, such as recommending friends, relevantinformation, goods and so on according to users’ potential interest. Nowadays a lotof popular social networking sites, such as Pengyou, Renren, Sina, also provide thefunction of recommendation, but the recommendation of these social networkingsites did not involve users’ text information, which makes some recommendationsmay not be able to meet users’ heart. Therefore, if can dig out microblogging users’potential interest accurately based on users’ text information, it can provide betterrecommended services to users. The potential interest of users can be not only usedfor recommended services, but also can be used for recommending ads, businessescan put appropriate ad to users based on users’ potential interest, this can get betterpublicity and commercial profits.The main contributions of this research are described as follows. Firstly, duringthe process of extracting candidate words, apart from the use of frequency_basedkeyword extraction method, also taking into account the position information of thewords. Secondly, theme model applied to the problem, and taking into account thecharacteristics of microblogging information and the requirements of theexperiment, the experiment uses a reduced Twitter_LDA to analysis potential topicsof candidate keywords. Thirdly, the SVM algorithm is used to analysis users’potential interest, to classify the extracted candidate keywords, in order toexcluding those who are non-interest term. Fourthly, in order to further understandmicroblogging users’ heart, also make sentiment analysis. Experimental resultsshow: Firstly, the adding of position information to candidate keywords, to a certainextent, improve the correct rate and the recall rate of the result of extractingpotential interest; Secondly, reduced Twitter_LDA some extent can improve thecorrect rate and recall rate of the result of extracting potential interest; Third, SVMalgorithm is perfectly suitable for the issue of microblogging users’ potential interest. Fourthly, different combinations of features can affect the final results.