A genetic algorithm based clustering using geodesic distance measure

Aim at the problem that classical Euclidean distance metric cannot generate a appropriate partition for data lying in a manifold, a genetic algorithm based clustering method using geodesic distance measure is put forward. In this study, a prototype-based genetic representation is utilized, where each chromosome is a sequence of positive integer numbers that represent the k-medoids. Additionally, a geodesic distance based proximity measures is adopted to measure the similarity among data points. Experimental results on eight benchmark synthetic datasets with different manifold structure demonstrate the effectiveness of the algorithm as a clustering technique. Compared with generic k-means algorithm for clustering task, the presented algorithm has the ability to identify complicated non-convex clusters and its clustering performance is clearly better than that of the k-means algorithm for complex manifold structures.