Author : Dr. Sangeeta Ahuja 1
Date of Publication :7th December 2016
Abstract: Germplam Evaluation plays very important role in genetic resources management and hybrid selection. The germplasm evaluation using the Kernel Density based Bayesian Ensemble (KDBCE) method gives promising results. It is based on Multivariate Kernel Density Estimation and Naive Bayes Classifier to obtain robust clustering using Density based (DBSCAN) clusterer. This algorithm operates in three phases. During the first phase, H input clustering schemes are generated by using the density based algorithm (DBSCAN) with different number of clusters in each clustering scheme. The optimum number of clusters is determined by computing the Silhouette coefficient for each clustering scheme. The second phase equalizes the number of clusters generated by different clustering schemes depending upon the optimum number of clusters. Accordingly, the clusters split or merge in different clustering schemes by using the kernel density based split and merge method. In the third phase, consensus partition is generated by the Naive Bayes Classifier. Empirical evaluation of the algorithm shows that the proposed method significantly improves the quality of resultant clustering scheme compared to the best of the original schemes.
Reference :