A formal approach to chart patterns classification in financial time series

Classifying chart patterns from input subsequences is a crucial pre-processing step in technical analysis. In this paper, we compile comprehensive formal specifications of 53 chart patterns reported in the literature. A first-order logic representation is chosen to describe the shape and corresponding constraints of each pattern. These formal specifications are formulated in such a way that data mining algorithms can use them for classification without significant modification. These formal specifications are also intended to serve as a reference model for future research in the chart patterns classification area. Using these formal specifications, we perform extensive experiments using real datasets from NYSE Composite (NYSE), Hang Seng Index (HSI), and Amazon (AMZN). The performance of the proposed method is compared against Template Based (TB), Euclidean Distance (ED), and Dynamic Time Warping (DTW) approaches. The experimental results show that the rules translated from the specifications can be effectively used to identify chart patterns from real datasets.