A memetic-based technical indicator portfolio and parameters optimization approach for finding trading signals to construct transaction robot in smart city era
Affiliations: [a] Department of Computer Science and Information Engineer, National Kaohsiung University of Science and Technology, Kaohsiung, Taiwan | [b] Department of Information and Finance Management, National Taipei University of Technology, Taipei, Taiwan | [c] Department of Computer Science and Information Engineer, Tamkang University, Taipei, Taiwan
Correspondence:
[*]
Corresponding authors: C.H. Chen, Department of Computer Science and Information Engineer, National Kaohsiung University of Science and Technology, Kaohsiung, Taiwan. E-mail: chench@nkust.edu.tw. R.D. Chiang, Department of Computer Science and Information Engineer, Tamkang University, Taipei, Taiwan. E-mail: 081863@mail.tku.edu.tw.
Abstract: With the development of smart cities, the demand for personal financial services is becoming more and more importance, and personal investment suggestion is one of them. A common way to reach the goal is using a technical indicator to form trading strategy to find trading signals as trading suggestion. However, using only a technical indicator has its limitations, a technical indicator portfolio is further utilized to generate trading signals for achieving risk aversion. To provide a more reliable trading signals, in this paper, we propose an optimization algorithm for obtaining a technical indicator portfolio and its parameters for predicting trends of target stock by using the memetic algorithm. In the proposed approach, the genetic algorithm (GA) and simulated annealing (SA) algorithm are utilized for global and local search. In global search, a technical indicator portfolio and its parameters are first encoded into a chromosome using a bit string and real numbers. Then, the initial population is generated based on the encoding scheme. Fitness value of a chromosome is evaluated by the return and risk according to the generated trading signals. In local search, SA is employed to tune parameters of indicators in chromosomes. After that, the genetic operators are continue employed to generate new offspring. Finally, the chromosome with the highest fitness value could be provided to construct transaction robot for making investment plans in smart city environment. Experiments on three real datasets with different trends were made to show the effectiveness of the proposed approach, including uptrend, consolidation, and downtrend. The total returns of them on testing datasets are 26.53% 33.48%, and 9.7% that indicate the proposed approach can not only reach risk aversion in downtrends but also have good returns in others.
