Prediction Of Hot Micro-Blog Forwarding Amount Based On The Personality Analysis

The impact of micro-blog has become one of the hot spots in the social network research, and it has important research value in the monitoring of public opinion, marketing, and communication and so on. A micro-blog’s forwarding and comments mark the public’s concern and participation in this article, it has also become a measure of the spread of news, the impact of the event. At present, most study of micro-blog is from the network and social angles, focusing on three aspects:1) Individual forward intention research:the research on the willingness of a single user, including whether it is willing to forward and the factors that affect the willingness of forwarding.2) Forwarding impact factors research:the factors that affect the user’s forwarding, including the use of micro-blog’s behavior and social scope.3) Forwarding behavior prediction:analysis whether the user will forward a micro-blog that related to a particular event. But these studies are limited to a particular event, or for a particular range of people.In this paper, we will try to predict hot micro-blog forwarding amount based on the personality analysis. The contributions of this paper are as follows:1. Introduce the "Big Five" theory and its role in the micro-blog text analysis, and established the "Big Five" dictionary based on Chinese LIWC dictionary and emotion dictionary. Users can input text to get "Big Five" score by using "Big Five" dictionary.2. Analyzing micro-blog’s text, and calculating micro-blog’s personality score. Personality score of the micro-blog shows five dimensions. The score will be used to train the relationship between personality and forwarding.3. Train the relationship between personality and forwarding and predict forwarding. In this paper, we use three kinds of training methods:multiple linear regression based on least square method as the basic method, local weighted regression and BP neural network as the improved contrast method. These three methods are used to simulate the relationship between personality and forwarding. The experimental results show that the local weighted regression has better prediction performance.