多目标进化算法在互联网基金理财产品投资组合中的应用研究

发布时间：2018-03-06 23:12

本文选题：多目标优化　切入点：多目标进化算法　出处：《华南理工大学》2015年硕士论文　论文类型：学位论文

【摘要】：近年来,随着移动互联网、云计算等技术的发展,互联网金融理财越来越受到人们的关注,各种基金理财产品可以很方便地通过网上平台、移动app应用购买。目前市场存在着多种基金理财产品,对于个人投资者以及互联网金融理财平台,如何在众多的基金产品中选择最优的投资组合,具有十分重要意义。基金理财本质上是一种投资组合问题,追求风险和收益两个重要目标,是典型的多目标优化问题,实际问题中基金种类较多,有很多附加限制例如交易费用、投资种类等,计算量庞大。多目标进化算法是一类可以有效求解复杂组合优化问题的算法,被广泛地应用于投资组合等实际问题。本文基于经典的投资组合均值方差模型,提出了新的改进模型,并根据实际问题,对NSGA-II算法的初始数据处理方式、遗传交叉和变异算子进行了改进。实验表明,这些改进显著地提高了算法的计算效率,并改善了应用效果。本文主要工作和创新点如下:1.以现代投资组合理论为基础,在经典的单目标均值方差模型上,考虑了交易费用、投资种类、购买价格水平等因素,提出了改进的多目标优化模型,该模型适合解决实际问题;2.重点查阅和研究了典型的多目标进化算法——NSGA-II,实现了该算法对模型的求解,同时通过实验比较了它与其它进化算法的收敛性和时间性能;3.针对NSGA-II算法在求解基金投资组合实际问题时时间过长、收敛性不够理想的问题,提出了有效的改进方法,首先改进初始数据的预处理方式,预先进行非支配选择,缩小了问题决策空间;然后采用二进制编码方式代替标准算法的实数编码方式,提出改进的交叉变异算子,定义了辅助可行解修复操作,满足改进模型的约束条件,降低了模型的求解复杂度。这些改进显著地提升了计算效率,改善了应用效果。4.为验证算法的应用效果,本文采用了两个数据集,一个数据集是标准数据集,用于验证多目标进化算法对求解投资组合问题的有效性;另一个数据集是网上爬取的实际基金数据,并进行了对比实验与结果分析,充分验证模型和算法改进的有效性。本文详细研究了基金投资组合问题,提出了实际有效的改进多目标优化模型;通过编程爬取国内真实的基金数据信息,最终实现了多目标进化算法对模型的优化求解。此外还针对标准算法计算效率与应用效果较差的缺点,提出了有效的改进方法。因此,本文的研究工作对于拓展多目标进化算法的应用、对于个人投资者和基金理财平台选取最佳投资组合有一定的理论和实际意义。
[Abstract]:In recent years, with the development of mobile Internet, cloud computing and other technologies, people pay more and more attention to Internet financial management. At present, there are many kinds of fund management products in the market. For individual investors and Internet financial management platform, how to choose the best portfolio among the many fund products, Fund financing is essentially a kind of portfolio problem, which pursues two important objectives of risk and income, and is a typical multi-objective optimization problem. In the practical problems, there are many kinds of funds. There are many additional restrictions, such as transaction costs, types of investment, and so on, and the computation is huge. Multi-objective evolutionary algorithm is a kind of algorithm that can effectively solve complex combinatorial optimization problems. Based on the classical portfolio mean variance model, a new improved model is proposed. According to the actual problem, the initial data processing method of NSGA-II algorithm is discussed. The genetic crossover and mutation operators have been improved. Experiments show that these improvements have significantly improved the computational efficiency of the algorithm and improved the effect of application. The main work and innovations of this paper are as follows: 1. Based on the modern portfolio theory, Based on the classical single-objective mean variance model, an improved multi-objective optimization model is proposed, which takes into account the transaction cost, investment type, purchase price level and so on. The model is suitable for solving practical problems. The typical multi-objective evolutionary algorithm NSGA-IIs is studied, and the model is solved by this algorithm. At the same time, the convergence and time performance of the algorithm and other evolutionary algorithms are compared by experiments. Aiming at the problem that the NSGA-II algorithm takes too long to solve the actual problem of fund portfolio, and the convergence is not ideal, an effective improvement method is proposed. Firstly, the preprocessing method of the initial data is improved, and the non-dominant selection is carried out in advance, which reduces the decision space of the problem, and then the binary coding method is used to replace the real number coding method of the standard algorithm, and an improved crossover mutation operator is proposed. The auxiliary feasible solution repair operation is defined, the constraint condition of the improved model is satisfied, and the complexity of solving the model is reduced. These improvements have significantly improved the computational efficiency and improved the application effect .4.In order to verify the application effect of the algorithm, In this paper, two data sets are used, one is the standard data set, which is used to verify the effectiveness of the multi-objective evolutionary algorithm for solving portfolio problems, the other is the actual fund data crawling on the net. The comparison experiment and result analysis are carried out to verify the effectiveness of the improved model and algorithm. In this paper, the fund portfolio problem is studied in detail, and a practical and effective improved multi-objective optimization model is proposed. Finally, the optimization solution of the model by multi-objective evolutionary algorithm is realized by crawling the real fund data information in China. In addition, an effective improvement method is proposed to solve the problem of poor computational efficiency and application effect of the standard algorithm. The research work in this paper is of theoretical and practical significance to expand the application of multi-objective evolutionary algorithm and to select the best portfolio for individual investors and fund management platforms.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：F724.6;F832.2;TP301.6

【参考文献】