Mining Stock Price Changes for Profitable Trade Using Candlestick Chart Patterns

Abstract

One major technical analysis of stock price fluctuation is the use of candlestick charts. This paper proposes a model with six parameters to retrieve similar candlestick patters to improve accuracy of stock price predictions. Because criteria that trigger reversing trade largely affect gains and losses, we examine two criteria; one based on sum of negative stock price changes and the other on sum of negative 5-day average differences. The proposed retrieval algorithm and criteria are evaluated through simulations in terms of gains and losses using NASDAQ's daily stock data. The results of simulations indicate that the proposed method leads to a trade decision that opportunities of successful stock trades are effectively above that of failure ones with several percentage of gains. Simulations also show that high risks deliver high returns. The results are examined statistically by the regression analysis suggesting the significant capabilities of the proposed method to predict stock price fluctuations.