Candlestick chart

Abstract: Forecasting stock prices plays an important role in setting a trading strategy or determining the appropriate timing for buying or selling a stock. We propose a model, called the feature fusion long short-term memory-convolutional neural network (LSTM-CNN) model, that combines features learned from different representations of the same data, namely, stock time series and stock chart images, to predict stock prices. The proposed model is composed of LSTM and a CNN, which are utilized for extracting temporal features and image features. We measure the performance of the proposed model relative to those of single models (CNN and LSTM) using SPDR S&P 500 ETF data. Our feature fusion LSTM-CNN model outperforms the single models in predicting stock prices. In addition, we discover that a candlestick chart is the most appropriate stock chart image to use to forecast stock prices. Thus, this study shows that prediction error can be efficiently reduced by using a combination of temporal and image features from the same data rather than using these features separately.

Abstract: We conduct the first robust study of the oldest known form of technical analysis, candlestick charting. Candlestick technical analysis is a short-term timing technique that generates signals based on the relationship between open, high, low, and close prices. Using an extension of the bootstrap methodology, which allows for the generation of random open, high, low and close prices, we find that candlestick trading strategies do not have value for Dow Jones Industrial Average (DJIA) stocks. This is further evidence that this market is informationally efficient.

Abstract: In this paper, two hybrid models are used for timing of the stock markets on the basis of the technical analysis of Japanese Candlestick by Support Vector Machine (SVM) and Heuristic Algorithms of Imperialist Competition and Genetic. In the first model, SVM and Imperialist Competition Algorithm (ICA) are developed for stock market timing in which ICA is used to optimize the SVM parameters. In the second model, SVM is used with Genetic Algorithm (GA) where GA is used for feature selection in addition to SVM parameters optimization. Here the two approaches, Raw-based and Signal-based are devised on the basis of the literature to generate the input data of the model. For a comparison, the Hit Rate is considered as the percentage of correct predictions for periods of 1–6 day. The results show that SVM-ICA performance is better than SVM-GA and most importantly the feed-forward static neural network of the literature as the standard one.