Features extraction in the copper futures market of China based on DIV clustering method of interval data

In futures market, daily trading records of numerous contracts compose a large scale database. How to integrate the mass data, extract the general features, and find out the underlying operational regularity of the market are significant to instruct the sound development of the market. For that research motivation, this paper utilizes the symbolic data analysis (SDA) methodology to model on the large scale database of the copper futures contracts in Shanghai Futures Exchange (SHFE). First, classify the mass and complex time-series trading records in the temporal dimension as their residual time to maturity, so that the original data set of single-values is transformed into the high-level structure of symbolic interval data which greatly reduces the dimension scale of the sample space. Based on that, symbolic DIV clustering method is applied to the interval data and three clusters partitioned by the size of trading volume are obtained. Moreover, ZoomStar graphs of the three clusters are also clearly illustrated the variation features of the copper futures. The results of the empirical study indicate that in the whole valid period of transaction, speculations present a “less-more-less” quantity in trading volume and open interest, and the tradings in two and three months to maturity are especially more active than in the other time. The application research also verifies the validity and practicability of SDA in integrating complex system, effectively modeling and information mining.