Predicting the direction of Indonesian stock price movement using support vector machines and fuzzy Kernel C-Means

The nature of stock price fluctuations becomes a challenge for the investors to gain return in investing stocks. To overcome this problem, investors need some accurate predictions in order to anticipate future stock price movement. However, predicting the direction of stock price movement is a complex task due to many uncertain factors affecting the stock price itself. Therefore, this paper studied the application of Support Vector Machines and Fuzzy Kernel C-Means in predicting the direction of stock price movement of Indonesian stock market, particularly on banking subsector. Using the stock historical data, eight technical indicators have been computed to obtain two different approaches for the input model. One of them use the computed indicators while the other process the computed indicators into trends. The results suggest that, in general view, Support Vector Machines with technical indicators represented as trend being the input model outperforms the other prediction models. However, in particular condition, the best model of the entire observation with 92 % accuracy is given by FKCM with computed indicators as the input by using σ = 100 and 90 % training data.