Читать книгу Data Analytics in Bioinformatics - Группа авторов - Страница 47

This algorithm is based on dynamic modeling which is executed in two stages. In the first stage graph partitioning algorithm is applied resulting into smaller clusters. As graph partitioning technique is applied the elements in the clusters are strongly connected. This establishes a quality separation among the clusters by discarding the noisy data. The later stage is executed with the application of agglomerative hierarchical clustering combining all the smaller clusters iteratively. This algorithm doesn’t fall short of incorrect merging because of its 2 stage execution. And also CHAMELEON overcomes all the shortcomings of AGNES and DIANA in which the data once split cannot be merged [21].