DMT-TAFM: A Data Mining Tool for Technical Analysis of Futures Market

摘要

Technical analysis of financial markets describes many patterns of market behavior. For practical use, all these descriptions need to be adjusted for each particular trading session. In this paper, we develop a "Data Mining Tool for Technical Analysis of the Futures Markets" (DMT-TAFM), which dynamically generates rules based on the notion of the price pattern similarity. The tool consists of three main components. The first component provides visualization of data series on a chart with different ranges, scales, and chart sizes and types. The second component constructs pattern descriptions using sets of polynomials. The third component specifies the training set for mining, defines the "similarity" notion, and searches for a set of similar patterns. DMT-TAFM is useful to prepare the data, and then reveal and systemize statistical information about "similar" patterns found in any type of historical rice series. We performed experiments with our tool on three decades of trading data for hundred types of futures. Our results for this data set shows that, we can prove or disprove many "well-known" patterns based on real data, as well as reveal new ones, and use the set of relatively consistent patterns found during data mining for developing better futures trading strategies.