Relationship discovery in public opinion and actual behavior for social media stock data space

Abstract

With the rise of social data media, the cyber world nearly parallels to the real world. The trajectory of a hot event is reflected in social media by Public Opinion Data Space (OS) and Actual Behavior Data Space (BS). However, the relationships with a variety of mechanisms in each space or between them are often unknown. To solve the above issues, the traditional methods for inferring relationship are by performing a statistical similarity analysis of time sequence from dynamic elements. In specially, the research of clustering nonlinear correlation data object is rare, so we propose Matrix Similarity Clustering Algorithm (MSCA) based on random matrix theory and combined with sliding window technology to cluster the similarity of multidimensional time sequences. This method is effective to detect the trend relationship of time sequences with multiple dynamic elements. In addition, we construct a knowledge map to analyse the relationships in OS and BS.