摘要
统计数据轨迹一般具有重视变化趋势、数据噪声较大、模式分布不同等特点,直接使用传统的聚类分析方法难有很好的效果。对此在K-means算法的基础上,分别采用了归一化处理、平滑处理以及关键峰匹配等方法处理上述三个问题,设计了一种解决系统使用轨迹模式分析问题的改进聚类方法。通过使用仿真数据与实际数据进行测试分析,在仿真数据上改进算法显著降低了聚类的错误率。在实际数据上,改进算法得出的聚类结果优于K-means算法,由此证明了改进方法比传统K-means聚类算法在该问题上效果更好。
The traditional clustering analysis method has weaknesses for analysis of pattern clustering statistical data trajecto- ries, such as too much emphasis on changing trends, larger data noise, different distribution patterns and so on. Therefore, this paper proposed a method that based on K-means to optimize the functions used in clustering analysis. Before modeling, it eliminated the error of data sample and gave the experimental data normalized treatment, smoothing processing and critical peak matching method. Artificial and real data processing results show that the proposed method performs better than classic clustering method.
出处
《计算机应用研究》
CSCD
北大核心
2013年第10期3001-3006,共6页
Application Research of Computers