The Time-space Domain Of Trading Behavior Analysis Based On Event

Social network is different from ordinary static network, the network structure of social network change and changing with time. To some extent, Social network reflects the real world situation. Stock trading volume can well reflect the reality reaction of people who confront the significant events. The number of postings in stocks tieba and post content can well reflect the real reflection of groups in the virtual network world. However, the reflection of the groups in the virtual world and real world reflect might or might not agree. So, In this paper we also research whether exist consistency between the two world base on the event at last.This paper mainly studies the trading behavior based on events in the real world by using data mining technology to analyze stock trading daily data. Because of financial time series is non-stationary and nonlinear characteristics, so it must be divided into line in first step. This paper adopted a special method to show sequences based on the specific extremum points. Then we begin to use the anomaly inspection, clustering analysis to analysis sequence. Anomaly detection method is according to the characteristics of the stock trading volume, mainly depend on the properties of slope of the line. Using the method to find period(or points) of each stock trading time series respectively, we can also get the time domain of abnormal time domain at the same time. And each stock’s abnormal sequences are represented by event functions. Each stock’s anomalies are represented using an event function. Due to the stock trading time series is scalable, clustering analysis is the improved clustering method which based on dynamic time warping(DTW).Regard each abnormal time series of stock trading as an event, and use clustering analysis to find out related stocks which with the same abnormal transaction behavior based on an event, the stock’s field can be found in the knowledge map which researched by our laboratory, finally determine the time domain and space domain of the event in stock’s filed.In this paper, the research on airspace of trading behavior is to analysis the relation of the behavior of virtual world and the behavior of the real world when an event happed. In the experimental part, we also analyze the post’s number in each stock tieba, it belongs to the virtual world behavior research. Then finally we find out the stock tieba where investors post messages and discuss about hot topics which caused by an event. This result can be compared with the conclusion of the analysis of stock trading, and to judge the connection of the event’s influence on the two words.