Abstract: With the rapid growth of Weibo, its vast data mining technology has become a major academic topic in recent years. This paper provides a method of time-series prediction on stock market based on Weibo search and support vector machines (SVM), aiming at the application of Weibo data mining in the financial area. Topics, future trends and sentiments can be achieved three-level-classification on the basis of Weibo search, which can monitor investors’ sentiments on Weibo and calculate related sentiment indexes, namely bullish sentiment index (BSI (1)) and bearish sentiment index (BSI (2)). The multi-variable time-series prediction based on history data and BSI (MTPH&BSI) has formed, attributed to the introduction of the two indexes to the single-variable time-series prediction based on history data. After training patterns, optimizing parameters and testing patterns of prediction, the experimental results show that the proposed prediction model is better than the traditional model in predicting performance and generalizing abilities. This paper is significant to study the service capabilities of socialized media as Weibo.

Key words: support vector machines, Weibo search, stock market prediction, investor sentiment