论文部分内容阅读
电力客户分类应用领域广泛涉及数据挖掘与特征提取问题,为了提高聚类算法的稳定性和准确性,提出一种基于纵横交叉(Crisscross Optimization,CSO)算法的聚类方法,能有效克服k均值聚类算法对初始质心敏感,容易陷入局部极值的缺点。CSO算法采用一种双交叉搜索机制,其中横向交叉引入扩展因子增强全局搜索能力,纵向交叉引入维交叉概念,从而避免维局部最优问题。两种交叉算子交替产生中庸解,通过与父代竞争产生的占优解在种群中相互催化,从而避免早熟问题的同时能够迅速收敛到全局最优。利用新方法对电力大客户数据进行客观、科学的挖掘分析,实现了对电力大客户较全面和准确的精细化分类,为供电企业制定有针对性的营销策略提供了依据。
In order to improve the stability and accuracy of clustering algorithm, the application of power customer classification is widely involved in data mining and feature extraction. This paper proposes a clustering method based on Crisscross Optimization (CSO) algorithm, which can effectively overcome the problem of k- The class algorithm is sensitive to the initial centroid, easy to fall into the local extreme shortcomings. The CSO algorithm uses a two-cross search mechanism, in which the cross-directional cross-introduce expansion factor enhances the global search capability, and the vertical cross introduces the concept of dimension cross, so as to avoid the local optimization problem. The two crossover operators produce a neutral solution alternately, and the dominant solution produced by competing with the parent catalyze each other in the population, so as to avoid premature problems and quickly converge to the global optimum. Utilizing the new method to conduct an objective and scientific excavation and analysis of big customer data, a more comprehensive and accurate classification of large electric customers is achieved, which provides a basis for the power supply enterprises to formulate a targeted marketing strategy.