基于粗糙集的改进Leader聚类算法

An improved rough-based Leader clustering algorithm

传统的聚类算法,如Leader算法和k-Means方法等,只能处理完整精确的数据集,数据项只能被划分到一个簇.而粗糙集理论用上近似集和下近似集表示一个类,尤其当数据有噪声、不完全和不精确性时,非常有优势.在经典的Leader算法中引入粗糙集理论,以处理模糊数据的聚类,得到改进的Leader算法——IRL(improved roughbased Leader)算法.IRL算法首先扫描数据项集,生成初始L集、RL集、RU集;然后优化RU集;最后再合并L集、RL集、RU集,得到最后的聚类结果.实验结果表明,IRL算法非常有效.

Objects are partitioned into clusters with crisp boundaries in the conventional algorithms such as Leader algorithm and k-Means algorithm.However,rough set is represented with lower-bound and upper-bound,and is good for the case when the data is incomplete,inaccurate and noisy.In this paper,IRL（improved rough-based Leader）algorithm is proposed based on rough set and Leader algorithm.At first,the data set is scanned in order to gain the set L,RLand RU.And then,the set RUis optimized.At last,the set L,RLand RU are merged in order to find the clustering result.The experimental results show that the algorithm is effective.