基于遗传算法的程序化交易策略-应用数学专业论文.docxVIP

万方数据 万方数据 摘要 基于遗传算法的程序化交易策略 摘 要 遗传算法是一种优化仿生的随机搜索算法，模仿了自然界中生物进化的方法来解决 数学问题中的最优化问题。遗传算法具有以决策变量的编码作为运算对象和直接以目标 函数值作为搜索信息避开函数求导问题等优点，并且随着计算机技术的发展和 MATLAB、 SAS 等软件的不断完善，遗传算法开始得到越来越多国内外人士的关注，并广泛应用于 程序化生成、人工神经网络训练、机器学习等一系列超大规模不连续、非线性、多峰函 数的优化问题。 与一些经典数学方法相比，遗传算法还属于一类新兴学科，许多地方还存在问题等 待学者们从理论以及实现方法上不断的完善。在对遗传算法的原理以及国内外学者的研 究等进行详细分析后，本文针对基本遗传算法存在的缺点提出了相应的改进：（1）引入 了新的适应度函数，利用一个平滑适应度函数的数学技巧来保留适应度更高的个体，平 滑掉适应度低的个体来减少那些只能取得局部极值的个体，提高找到全局最优解的效 率；（2）对 Srinivas 提出的自适应遗传算法中 Pc 和 Pm 自适应公式的交叉和变异概率进 行了改进，引入了局部和全局交叉变异的权重，进而更好的从全局来调整交叉和变异概 率来提高最优解的质量。最后利用改进的自适应遗传算法使用 MATLAB 来对期货交易指 标中的 MACD、BOLL 和 RSI 指标进行结合优化处理，建立了程序化交易模型，测试模型 结果表明改进的自适应遗传算法的收敛速度和解的最优性都优于自适应遗传算法。 关键词：平滑适应度函数；自适应遗传算法；交叉概率；变异概率；全局收敛 I 华北水利水电大学硕士学位论文 II ABSTRACT PROGRAM TRADING STRATEGY BASED GENETIC ALGORITHM ABSTRACT Genetic algorithm is a random search arithmetic based on bionic optimization and imitates the method of biological evolution which solves mathematical optimization probolems in nature. Those are big advantages that genetic algorithm regards the code of decision variables as a operand and the objective function value is used for the search information directly to avoid function derivation problem. With the developments of computer technology ,such as MATLAB、SAS and other software. The genetic algorithm gets more and more peoples attention. And it is widely used in program production, artificial neural network training, machine learning and a series of large scale which is not continuous, non-linear, multi-modal function optimization problem. Compared with some classical mathematical methods, genetic algorithm is a kind of new disciplines. There are many problems for scholars to realize perfect continuous from the theory and the method.Some improvements are proposed in this article after a detailed analysis on the principle of genetic algorithm from the domestic and foreign scholars research according to the existing problems of basic genetic algorithm : (1)Introduced a new fitness function. Using a m