Multi-objective Invasive Weed Optimization algortihm for clustering

In this paper, we proposed a new approach to solve the clustering problem in which the cluster number is uncertainty. It utilizes IWO (Invasive Weed Optimization) algorithm to optimize two fuzzy clustering objective function simultaneously, and a variable-length real-coded scheme has been adopted, the variable length weed encodes the cluster centers with variable numbers. In order to keep the diversity of the weeds, we introduce a new mechanism called feedback update mechanism to update the individuals which the corresponding number of cluster centers has been eliminated in one generation. Finally, the Silhouette index is used to select the best solution. The algorithm is used to cluster 15 artificial data sets and 4 real life data sets and shows good performance.