Hybrid Technical-Visual Features for Stock Prediction

Stock prediction is an interesting and important problem for analysts and investors. However, the factors affecting the movement of stock prices are very complicated, making it very difficult to effectively predict them. In the literature, technical analysis has been widely used for stock prediction, the analysis of candlestick charts being one of the main methods. In this paper, we apply an artificial intelligence technique based on the wavelet transform to automatically extract the visual features of candlestick charts. The prediction performance obtained using general technical indicators, the visual features of candlestick charts, and a combination of the two is examined. The experiments, based on data for the Taiwan stock market including ten electronics companies, show that combining both technical and visual features can allow different classifiers to achieve better prediction performance than the other methods.