将Dirichlet过程作为无穷高斯混合模型中权重参数的先验分布,利用贝叶斯定理得到参数的估计,并由Gibbs抽样算法得出聚类的个数和判断观测值的指示因子,利用统计模拟说明了算法的有效性,与传统方法相比,该方法误判率更低。%Dirichlet process is supposed as the prior distribution of the weights parameters based on infinite Gaussian mixture model. Bayes theorem is used to deduce the estimators of parameters. The Gibbs sampling algorithm is applied to obtain the clustering number and the judge indicator factor of the observation data. Finally, we conduct simulation to show that our algorithm can effectively cluster data and our algorithm have lower error rate than the traditional clustering methods.
展开▼