基于群体智能聚类研究及其在股市板块分析中的应用

发布时间：2018-01-22 23:09

本文关键词： 数据挖掘股市板块果蝇优化算法近邻传播聚类算法实时学习差异化距离　出处：《长春工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：大数据时代背景下,股票市场产生的数据越来越多,传统分析方法已很难满足现代对股市板块分析的需求,如何从海量股票数据中挖掘出有用信息,并对股市板块进行分析和预测成为了一个重要的研究课题。数据挖掘作为一种新的数据技术在此环境背景下迅速发展,为开发数据信息资源做出重大贡献。聚类是一种无监督学习的数据挖掘技术,可以充分地分析数据内部特征和数据间的联系。通过形成的多个簇,挖掘出数据之间潜在的联系与特征,避免直面庞大的数据集。群体智能算法是一种演化计算技术,具有实现简单、收敛速度快、全局收敛性优等优点,已受到越来越多国内外专家和学者的关注。群体智能优化算法是人工智能的一个重要分支,与人工生命联系紧密。通过模拟自然界生物群体的各种行为,利用个体之间的信息传递和合作实现寻优的目的。果蝇优化算法(Fruit Fly Optimization Algorithm,FOA)是群体智能优化算法的一种,因其寻优精度高且设置参数少,已经广泛应用到多个领域。但是FOA仍然存在不足:FOA在处理复杂函数问题时,收敛速度减慢,且收敛精度降低。针对此问题提出一种实时学习的果蝇优化算法(A Real-Time Learning Fruit Fly Optimization Algorithm,RTLFOA)。RTLFOA通过实时的获取种群知识,在种群陷入早熟时,通过获取的种群经验引导果蝇群体进行不同尺度的变异。变异后,味道浓度较差位置处的果蝇个体进行全局搜索,提高算法的收敛速度;味道浓度较优位置处的果蝇个体进一步对局部区域进行搜索,提高收敛精度。通过比较RTLFOA与其它算法对基准函数收敛性的测试结果,证明了RTLFOA有效的提高了收敛速度和收敛精度。近邻传播聚类算法(Affinity Propagation,AP)是一种无监督学习的聚类算法,不必指定聚类数目且处理大规模数据集具有较好的聚类结果。但是AP聚类算法的偏向参数设置难度大。(1)基于群体智能优化算法的全局搜索能力,首先提出基于差异化距离的果蝇优化算法(Fruit Fly Optimization Algorithm based on Differential Distance,FOADD)。FOADD算法使距离原点相同距离处的果蝇同时移动,避免了无效移动,提高算法的搜索效率。(2)基于FOADD的快速搜索能力,提出基于差异化距离果蝇优化的近邻传播聚类算法(Affinity Propagation based on FOADD,AP-FOADD)。在搜索空间内对偏向参数进行寻优,并使用经典数据集进行测试,根据Silhouette有效性指标与聚类结果证明其有效性。基于AP-FOADD聚类算法,对股市板块进行分析。本文选取了112家北京上市公司,并以四项可以反映上市公司盈利能力的财务指标为研究对象,对上市公司聚类。通过聚类结果对上市公司的经营能力、发展潜力进行分析,并根据上市公司年度报告证明聚类结果的有效性,为投资者提供投资组合、规避投资风险提供理论依据。
[Abstract]:Under the background of big data era, more and more data are produced in the stock market. The traditional analysis method has been difficult to meet the needs of the modern stock market plate analysis, how to extract useful information from the massive stock data. And the analysis and prediction of the stock market has become an important research topic. As a new data technology, data mining is developing rapidly in this environment. Clustering is an unsupervised learning data mining technology, which can fully analyze the relationship between the internal characteristics of data and data. Through the formation of multiple clusters. In order to avoid facing the huge data set, swarm intelligence algorithm is a kind of evolutionary computing technology, which has the advantages of simple implementation, fast convergence speed and excellent global convergence. Swarm intelligence optimization algorithm is an important branch of artificial intelligence, which is closely related to artificial life. Fruit Fly Optimization Algorithm was optimized by using the information transmission and cooperation among individuals. FOAis a kind of swarm intelligence optimization algorithm, which has been widely used in many fields because of its high precision and low parameter setting. However, FOA still has shortcomings in dealing with complex function problems. The rate of convergence slows down. And the convergence accuracy is reduced. A real-time learning algorithm for Drosophila is proposed. A Real-Time Learning Fruit Fly Optimization Algorithm. RTLFOA).RTLFOA acquires population knowledge in real time. When the population falls into early maturity, the acquired population experience leads Drosophila populations to mutate on different scales. In order to improve the convergence rate of the algorithm, the individual of Drosophila melanogaster at the location of poor taste concentration is searched globally. In order to improve the convergence accuracy, the individual of Drosophila melanogaster at the optimal location of flavor concentration further searched the local area to improve the convergence accuracy. By comparing the results of RTLFOA and other algorithms to test the convergence of reference function. It is proved that RTLFOA can effectively improve the convergence speed and accuracy. The near neighbor propagation clustering algorithm is Affinity Propagation. App) is an unsupervised learning clustering algorithm. There is no need to specify the number of clustering and processing large data sets has better clustering results, but AP clustering algorithm is difficult to set the bias parameters.) based on the global search ability of swarm intelligence optimization algorithm. First of all, an optimization algorithm for Drosophila based on differential distance is proposed (. Fruit Fly Optimization Algorithm based on Differential Distance. FOADD).FOADD algorithm makes Drosophila at the same distance from origin to move at the same time, avoids invalid movement, and improves search efficiency. 2) Fast search ability based on FOADD. A neighbor propagation clustering algorithm, Affinity Propagation based on FOADD, is proposed based on the optimization of the differential distance of Drosophila melanogaster. AP-FOADD. to optimize the bias parameters in the search space, and use the classical data set to test. According to the Silhouette validity index and clustering results to prove its validity. Based on the AP-FOADD clustering algorithm, the stock market block is analyzed. In this paper, 112 listed companies in Beijing are selected. And taking four financial indexes which can reflect the profitability of listed companies as the research object, this paper analyzes the business ability and development potential of listed companies through clustering results. According to the annual report of listed companies, the validity of clustering results is proved, which provides a theoretical basis for investors to invest portfolio and avoid investment risk.
【学位授予单位】：长春工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F832.51;TP311.13

【相似文献】