基于粒子群优化的基因表达式编程分类算法研究与应用

发布时间：2018-04-02 15:39

本文选题：基因表达式编程　切入点：粒子群优化　出处：《浙江工业大学》2014年硕士论文

【摘要】：数据挖掘中的分类是当今计算机应用技术和理论研究中的热门领域,它作为一种有效的数据分析手段,具有广泛的应用。将进化计算与分类技术结合,形成基于进化计算的分类方法是其中一个重要的研究方向。基因表达式编程(Gene Expression Programming, GEP)和粒子群优化(Particle Swarm Optimization, PSO)算法是两种新型进化计算方法,它们分别通过模拟生物进化机制和鸟类觅食行为搜索问题的最优解。本文以GEP和PSO为工具,研究这两种进化计算方法在基于距离的分类方法中的应用。本文的主要工作和成果如下： 1.介绍了基因表达式编程和粒子群优化算法,包括算法的起源、基本流程和相关概念。在此基础上,比较了这两种进化计算方法的异同点。 2.详细阐述了基于距离的分类方法的基本原理。提出了基于GEP的类中心点分类算法,引入了一种新的运算符,研究了该算法中GEP个体的编码和解码问题。多个数据集上的实验证明基于GEP的类中心点分类算法具有较强的搜索能力,且分类效果较好。针对基于GEP的类中心点分类算法后期存在收敛速度变慢,易陷入局部最优值的情况,引入了PSO算法,提出了基于PSO的GEP分类算法,解决了该算法从GEP阶段转换到PSO阶段过程中两类种群个体的编码兼容问题。实验证明基于PSO的GEP分类算法陷入局部最优值的情况减少,且分类精度较高。 3.推广算法的应用领域,将基于PSO的GEP分类算法应用于面向对象的遥感图像分类,提出了基于GEPSO (GEP and PSO)模型的面向对象遥感图像分类方法。在加拿大安大略省伦敦市中部航空正射影像上的实验证明,基于GEPSO模型的面向对象遥感分类方法具有可行性
[Abstract]:Classification in data mining is a hot field in computer application technology and theoretical research. As an effective means of data analysis, it has a wide range of applications.It is an important research direction to combine evolutionary computing with classification technology to form a classification method based on evolutionary computing.Gene Expression programming (GP) and particle swarm optimization (PSO) are two new evolutionary algorithms, which simulate the evolutionary mechanism of organisms and the optimal solution of bird foraging behavior.In this paper, we use GEP and PSO as tools to study the application of these two evolutionary computing methods in distance-based classification.The main work and results of this paper are as follows:1.This paper introduces the genetic expression programming and particle swarm optimization algorithm, including the origin, basic flow and related concepts of the algorithm.On this basis, the similarities and differences between the two evolutionary computing methods are compared.2.The basic principle of distance-based classification method is described in detail.A class center point classification algorithm based on GEP is proposed, and a new operator is introduced. The coding and decoding problems of GEP individuals in this algorithm are studied.Experiments on multiple data sets show that the algorithm based on GEP has strong searching ability and good classification effect.In view of the situation that the convergence rate of the class center point classification algorithm based on GEP becomes slow and easy to fall into the local optimal value in the later stage, the PSO algorithm is introduced, and the GEP classification algorithm based on PSO is proposed.This algorithm solves the problem of coding compatibility between two classes of individuals in the transition from GEP stage to PSO stage.The experimental results show that the GEP classification algorithm based on PSO has less local optimal value and higher classification accuracy.3.In this paper, the GEP classification algorithm based on PSO is applied to object oriented remote sensing image classification, and an object oriented remote sensing image classification method based on GEPSO GEPSO and PSO model is proposed.The experiment on the aerial orthophoto image of the central city of London, Ontario, Canada proves that the object-oriented remote sensing classification method based on GEPSO model is feasible.
【学位授予单位】：浙江工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP311.13;TP751

【参考文献】