| Stock analysis and forecast has great realistic significance for listed companies and investors. Technical analysis calculate the rules to predict the future trend of the change of stock price based on historical data of stock trading. A technical indicators is also provides a technical trading rules, the early study of technical index analysis usually focus on one or several parameters of several trading rules to help decision-making, recent theory and research shows that the combination of the technical regulations can often gain more than adopting each rule separately.Portfolio selection is the cornerstone of modern financial decision theory, the theory of diversification research mostly focus on a variety of asset portfolio, research on the problem of portfolio investment always focus on single invest cycle, without considering the market changes. Despite of single phase of the portfolio and the study of multiphase portfolio problems has made certain achievements, yet the multi-periods(short, medium and long period) portfolio optimization problem research has just started, which was still in the stage of exploration.In view of the above facts, this study expand the technical rules to a variety of technical indicators with different time parameter, and join the combination of various complex trading strategies, with the technical analysis and data mining technology, this paper build the transaction rule mining model based on biclustering technique. Furthermore, we analyze the theory of diversification and the multi-period portfolio problems, quantitative each invest period in the process of portfolio investment of the capital allocation. This paper reseach the multi-period investment trading strategies based on biclustering, specific research work is as follows:(1) A detailed overview of stock analysis content and technology.(2) On the basis of technology analysis and data mining technology, we build a trading rules mining algorithm and model based on biclustering technique.Using the nested particle swarm algorithm to optimize the model parameters and the multi-period portfolio allocations.By comparing the models(BIC-NPSO) and other four algorithms as well as traditional Buy and Hold(Buy-and-Hold, BAH), the superiority of our model was confirmed by different stock data.In this paper, the main innovation point are as follows:(1) It is the first time to use biclustering algorithm to the design of the financial trading system. Consistent with the efficient design of biclustering algorithm to extract constant column in the process of price change to obtain trading signals, and classifying trading signals using KNN classification algorithm, and to obtain the intelligent trading system, and the effectiveness of the model is obtained by empirical research evidence.(2) This paper innovatively investigate and research the multi-period portfolio problem. This paper analyze the theory of diversification and the multi-period portfolio problems, quantitative each invest period in the process of portfolio investment of the capital allocation based on the change of stock price and to avoid high-risk and obtain high profits.(3) This paper put forward a two layer nesting particle swarm optimization algorithm. In this paper, we using layer nesting particle swarm optimization algorithm based on the maximization objective function, and optimize the parameters in the model of multi-period portfolio allocations, that is, the outer particle swarm algorithm of biclustering algorithm optimize two threshold parameters, the inner particle swarm algorithm optimize multi-period(short, medium and long period) portfolio allocation proportion.(4) This paper has built a framework of model of good scalability. This paper builds the model of good expansibility, which is easy to expand to a wider variety of technical indicators, different parameters and different investment cycle length, which makes this model very flexible for the vast number of investors and versatility. |
PDF Full Text Request