Comparison of the classification efficiencies of K-means and Hierarchical clustering methods for candlestick component length on the world gold price candlestick chart

บทคัดย่อ

In this study, the classification efficiencies of K-means and Hierarchical clustering methods for candlestick component length on the world gold price candlestick chart were compared with the objective to reduce the ambiguity in the identification of candlestick size. The data used in this study consisted of the opening price, closing price, highest trading price and lowest trading price from the daily gold price in the world market of the United States during January 2, 2012 to April 30, 2021 for a total of 2386 days. The data was divided into 2 groups: training data and testing data. The experiment consisted of the calculation of length, ratio, mean and standard deviation and standardization. The data was then classified using K-means and hierarchical clustering methods which it was found that K-means clustering method resulted in 5 clusters, while hierarchical clustering method resulted in 3 clusters. The classification efficiencies for each component of candlestick chart based on CCI values in both training data and testing data were all higher than 70 for both clustering methods. This indicated that the classification was quite effective. However, K-means clustering method was more effective compared to hierarchical clustering method for gold price data in the world market.