Dynamic stock trading system based on Quantum-inspired Tabu Search algorithm

Many heuristic methods or evolutionary algorithms such as Genetic Algorithm (GA) and Genetic Programming (GP) are common approaches used in financial applications. Determining the best time to buy and sell in a stock market, and thereby maximizing the profit with lower risks are important issues in financial research. Recent researches have used trading rules based on technical analysis to address this problem. These rules can determine trading times by analyzing the value of technical indicators. In other words, we can make trading rules by analyzing the value of technical indicators. A simple example of a trading rule would be, if one technical indicator's value achieves the pre-defined value, then we can buy or sell stocks. A combination of trading rules would become a trading strategy. The process of making trading strategies can be formulated as a combinatorial optimization problem. In this paper, a novel method which can be applied to a trading system is proposed. First, the proposed system uses the Quantum-inspired Tabu Search (QTS) algorithm to find the optimal combination of trading rules. Second, it uses sliding window to avoid the major problem of over-fitting. The experiment results of earning profit show much better performance than other approaches. Especially, the proposed method outperforms Buy & Hold method which is a common benchmark in this field.