【作者】 刘宇；

【导师】 郑逢斌；

【作者基本信息】 河南大学 ， 应用数学， 2008， 硕士

【摘要】 数据聚类是一个正在蓬勃发展的领域,涉及数据挖掘、统计学、机器学习、空间数据库技术、商务信息等领域,可以说涉及了人类社会生活的方方面面。模糊聚类分析是将模糊理论应用到聚类分析中,为显示数据提供了模糊处理能力,在许多领域被广泛应用。FCM(Fuzzy c-means)算法是模糊聚类中的一种重要方法,它具有算法简单、局部搜索能力强且收敛速度快的特点,然而FCM算法受初始化影响较大,在迭代时容易陷入局部极小。遗传算法是一种随机搜索的全局优化算法,它以一个种群中的所有个体为对象,利用随机化技术指导,对一个被编码的参数空间进行高效搜索。其求解过程简单,是智能计算中的主要算法之一。将FCM算法和遗传算法相结合,将对算法的全局优化能力产生巨大作用,使算法性能大为提高。但简单遗传算法采用固定的交叉概率和变异概率,若直接采用简单遗传算法进行聚类,会出现收敛过慢,稳定性差等问题。本文对遗传算法的交叉概率和变异概率进行了深入研究,提出了一种新的交叉概率和变异概率,并将遗传算法和FCM算法相结合,提出了一种基于自适应遗传算法的模糊聚类算法(Adaptive Genetic Algorithm Fuzzy c-means,AGAFCM),该算法能充分发挥遗传算法的全局优化特征和FCM算法局部搜索能力,极大地提高了算法的精度和效率。绩效考核作为人力资源管理的一项基础性工作为人力资源管理提供准确的反馈信息。本文将改进的自适应遗传模糊聚类算法用于员工绩效考核成绩的聚类分析,构建了聚类分析的员工绩效考核模型。为现代企业人力资源管理的绩效考核提供了一种有效的数据分析方法。同时将该模型应用于人力资源管理系统,证明了模型的实用性和有效性。本文的主要工作和贡献如下:1)对遗传算法的交叉概率和变异概率进行改进,提出了根据个体适应度值进行线性调整的新的交叉概率和变异概率,以提高遗传算法的收敛速度和稳定性。2)采用改进的交叉概率和变异概率和一些已有的改进策略对遗传算法进行改进,同时将改进的遗传算法应用于模糊聚类,提出了一种基于自适应遗传算法的模糊聚类算法(AGAFCM)。实验证明,本文提出的算法在收敛速度、稳定性和聚类准确率方面都有明显改善。3)将本文提出的AGAFCM算法用于员工绩效考核成绩的聚类分析,构建了聚类分析的员工绩效考核模型,为人力资源管理者对企业整体人力资源水平的分析提供了准确的信息。更多还原

【Abstract】 Data clustering is an emerging area which involves various areas as data mining, statistics, machine learning, spatial database technology and business information, etc. Fuzzy cluster analysis put fuzzy theory into application of cluster analysis to provide capability in data display and is widely used in various areas. FCM(Fuzzy c-means) algorithm is one of important methods in fuzzy clustering, it has the characteristics as simple, fast convergence and strong local searching power, etc. However, FCM is sensitive to initialization and tends to result in local minimum in iterations.Genetic Algorithm is a random searching global optimization algorithm, it targets all the individuals of a population and searches effectively in a coded parameter space with the random technical guidance. Because of its simple solution procedure, it becomes one of the main algorithms in intelligence computing. The combination of FCM algorithm and genetic algorithm benefits the global optimization and makes tremendous improvement in algorithm performance. However, a simple genetic algorithm only uses a fixed-probability and the mutation rate for solution which has shortcomings like slow convergence and poor stability.This paper studies the crossover and mutation probability of generic algorithm and presents a new crossover and mutation probability. A new adaptive generic algorithm fuzzy c-means(AGAFCM) which combines both generic algorithm and FCM is proposed. AGAFCM takes full advantage of the global optimization of generic algorithm and the local search power of FCM and has great improvement in both accuracy and efficiency.Performance assement is a basic part of human resources management and provide accurate feedback for human resources management. The paer puts improved adaptive generic and fuzzy cluster alogrithm into cluster analysis of employee performance assement, generating an employee performance assement model for cluster analysis and provides an effective data analysing method for modern emterprize human resource management. The model is applicated in practical human resource management and definitely proves the usage and effectiveness.The paper presents the main works and contributions as follows: 1)Improves the crossover and mutation probability of generic algorithm. It provides new cross-over and mutation probability which is adjusted according to the individual adaptive ability so as to speed up the convergence and stabability.2)Improves the generic algorithm operation and put it into the application of fuzzy cluster and proposes a adaptive generic algorithm based fuzzy clustering. The algorithm presented in this paper is proved by exprements in convergence speed, stabability and accuracy.3)Created a new employee performance model by putting the new AGAFCM into use of cluster analysis for employee performance assement and provides human resouces mangement with accurate information in analysis to overall human resource in enterprise.更多还原