Automated trading based on biclustering mining and fuzzy modeling

More and more records or charts of historical financial data are used for technical analysis, hoping to identify patterns that can be exploited to achieve excess profits. Technical analysis has been widely used in the real stock market to forecast stock price or stock trading points. The good association of technical indicators can obtain good prediction results in stock markets. But the selection of technical indicators is also a tough problem. In this paper, we introduce a forecasting model incorporating biclustering algorithm with a new fuzzy inference method. Biclustering algorithm discover biclusters which are regarded as trading patterns. And a new fuzzy inference method is used for determining trading points. The proposed forecasting model (called BM-FM) was used for predicting three real-world stock data. The experiment is designed by comparing the profit ratio in TPP-based strategy, IPLR and IPLR-ANN with the profit ratio in our forecasting model. According to experimental results, it is indicated that our model obtains more earnings and higher profit ratio than other comparative methods.