Method for the analysis of the preferences of network users

At present, research on users’ preferences is becoming increasingly widespread. Traditional research used a simple cosine similarity algorithm to calculate the similarity between users and just focused on one factor. So, a more complete method is described for researching preferences by considering more factors of the network to analyse data. Two personalisation recommendation algorithms are given based on users’ preferences. These two algorithms mainly work on two aspects: one is the content-based recommendation algorithm. It uses a cosine similarity algorithm to calculate the similarity between users, uses LDA to obtain improvement and uses screens according to a user impact factor. The other is a relation-based recommendation algorithm. It uses the PersonalRank algorithm to calculate the similarity, and uses the weight of users’ communication behaviours to improve the algorithm. Finally, micro-blog users’ data are analysed in detail, and the consequences of the two recommendation methods are fused based on the content of the results and recommendations. As a result, we obtained more accurate recommendation results and verified the feasibility of the algorithm by experimental analysis.