Research On Predicting Stock Comovement Based On Microblog Attributes

Stock comovement refers to the interaction between stock prices of the same industry or sector in the stock market, mainly for the same up or down with. Therefore, stock return comovement analysis is a very important guidance to financial analysis, investment strategies, academic research, market risk assessment. In recent years, with the rapid development of network information technology, the emerging social softwares such as microblog open up a high-speed flow of information era. and provide a broad social platform for human communication and information sharing. Firms with microblog accounts can publish information and interact with microblog users directly to enables users to have access to information more easily, which has broken the block and asymmetry of the traditional media communication.Previous research has shown that firms with official microblog accounts have a much higher comovement than those without such accounts. We can classify the firms into new homogeneous groups by firm account's number of followers to predict the comovement of stocks.Based on the attributes of microblog account, we propose a more accurate classification scheme, the microblog attributes adopted by this scheme are the most significant indicators to reflect the microblog account's influence and make the stocks within the same group have a higher comovement. In the proposed scheme, we collect the attributes of firms with official microblog accounts and correspondent stock exchanges data, then we apply the K-means, an aggregation algorithm to classify firms into different stock groups and exam the pairwise stock return correlations within the same groups. Finally, we compare the results with previous classification method on the stock comovement.The results show that the proposed classification scheme in this paper is feasible. Stocks classified by the proposed microblog attributes have a higher comovement than those by the number of followers. The proposed attributes can not only predict the comovement of stocks but also notably increase the accuracy of comovement predicting, compared with industry categories.