An empirical comparison of four initialization methods for the K-Means algorithm

摘要: In this paper, we aim to compare empirically four initialization methods for the K-Means algorithm: random, Forgy, MacQueen and Kaufman. Although algorithm is known its robustness, it widely reported in literature that performance depends upon two key points: initial clustering instance order. We conduct a series of experiments draw up (in terms mean, maximum, minimum standard deviation) probability distribution square-error values final clusters returned by independently on any order when each used. The results our illustrate random Kaufman outperform rest compared as they make more effective independent addition, convergence speed using methods. Our suggest method induces desirable behaviour with respect than method.