| Clustering analysis algorithm is an unsupervised algorithm and is widely used in the field of data mining which is known for not need to be divided into the training sample and test sample sets and also not need to do any hypothesis to samples. There are a lot kinds of clustering algorithms, such as K-Means, density clustering and hierarchical clustering and so on. But most of the clustering algorithm has the disadvantages of getting the number of clusters or initializing the center of clusters in advance. However, it is an difficult problem to determine the clustering number in advance in the absence of any prior knowledge, and random initialization of the clustering center also contributes to the clustering results with volatility and instability. In particular, how to determine the optimal number of clusters adaptive is an important basic subject in the field of data mining which need us to give further study.The mainly content of this topic around to reduce error which is brought by random initialization of clustering center, evaluate the effectiveness of the clustering results, and determine the optimal number of clusters. Based on the analysis of the traditional method, this paper put forward an improved and adaptive fuzzy SOM algorithm to determine the optimal number of clusters. The clustering results of this method can keep it’s stability and is different from the method such as K-means which the clustering results are impacted by initializing the clustering center randomly. In order to get the optimal number of clusters and reduce the complexity of the calculation, we construct a new kind of coefficient indicator which based on the extreme boundary samples to evaluate the quality of cluster. According to the minimum value of the line chart about clustering quality, we can get the optimal number of the clusters.The experimental results show that the method presented in this paper can determine the optimal number of clusters effectively. |
PDF Full Text Request