Strategy Modeling and Classifier Training for Share Trading

Abstract

Abstract. In technical analysis, each trading strategy can trigger a buy or sell action whenever the specified conditions are satisfied. When a set of strategies are applied to a particular stock, a trader often receives conflicting recommendations from each strategy. In this paper, we propose a unified data mining approach in which the outcomes of each strategy are taken into consideration for decision making. First, we develop a framework for composing complex trading strategies. Next, we show how to perform simulation analysis on constructed strategies using extracted historical prices. The result of the simulation analysis is then used for training classifiers which can be used for recommending stock trading actions. Experiments conducted with the price data from Hong Kong Stock Market show promising results.