Home // International Journal On Advances in Software, volume 11, numbers 3 and 4, 2018 // View article
Authors:
Yoshihisa Udagawa
Keywords: Stock price prediction; Technical analysis; Candlestick charts; Longest common subsequence; Statistics of candlesticks.
Abstract:
The paper describes a method for a short-term stock price prediction based on candlestick chart techniques that are popular among stock traders using technical analysis. While the techniques have long history, there is still no consistent conclusion on the predictability of the techniques. We focus on the fact that a trend of stock prices often continues after intervals of several days because stock prices tend to fluctuate according to announcements of important economic indicators, economic and political news, etc. Typically, stock price movements in the period without important news are small, resulting in generating a series of noisy candlesticks. To cope with the noisy candlesticks, this paper focuses on a dynamic programming algorithm that allows us to perform partial matches on sequences of stock prices. We propose a model consisting of six parameters for retrieving similar candlestick charts in order to take into account where the stock price occurs in high/low price zones, in addition to a price change and a length of candlestick body. Experiments are performed on the daily NASDAQ composite index. We choose the daily time frame since important news that affects stock prices occurs on a daily basis. The statistics of the candlesticks are calculated to determine the parameter values of the proposed model based on the average and the standard deviation. Experimental results show that the proposed method is effective in predicting both uptrend and downtrend. Strictly, the prediction of the downtrend is a little bit difficult than that of the uptrend, probably reflecting the fact that the NASDAQ stock market is constantly growing.
Pages: 440 to 451
Copyright: Copyright (c) to authors, 2018. Used with permission.
Publication date: December 30, 2018
Published in: journal
ISSN: 1942-2628