基于密度检测的EM算法

EM algorithm based on density detection

期望最大化(expectation maximization,EM)算法在聚类过程中无法识别噪声点,最终的收敛效果也依赖于初始值的选择。提出了基于密度检测的EM算法(DDEM),通过基于密度的方法来检测噪声点,利用基于密度和距离的方法进行初始值选择,改善了EM算法收敛效果。实验结果表明新算法可有效识别噪声点,降低初始值选择对收敛效果的影响,明显提高了聚类准确率和稳定性。

Expectation maximization algorithm may not effective enough to detect noises and its convergence result also relies on the selection of initial value in the clustering process. This paper proposed a new algorithm to improve the convergence effort of EM algorithm, which detected noises and selects initial value based on density detection and distance computing. Experiments show that the proposed algorithm can improve the accuracy as well as the stability of clustering by effectively noises detection and reduction of the influence of initial value in the convergence effort.