Pattern Recognition in the Japanese Candlesticks

Abstract

Pattern recognition analysis based on k-nearest neighbors classifiers is applied to the representation of the stock market dynamics with the help of the Japanese candlesticks augmented by the accompanying volume of transactions. Examples from a post-emerging Warsaw stock market are given. Conditions under which the Japanese candlesticks appear to have a reasonable predictive power are provided. The dependence of the results on the number of nearest neighbors, the length of the candlestick sequence, and the forecast horizon are shown. Possible ways of the forecast improvement are discussed.