Multiple index improvement of K-means clustering based on learning rate automatically choice

The routine K-means clustering and its upgrading alternatives were found with some problems as these: deleting Dimension Trap, failure to identifying initially the clustering center and the number of clustering attributes. In regard to these problems, a new clustering method was discussed along with two parameters introduced assessing clustering effect. The new approach addressed these problems by adopting different methods. It deleted dimension traps by using learning rate for autonomous learning. The application of Distance Seeking based on Consistency and DB Index identified initially its center and the number of clustering units. Average of DB Index and DB quadratic mean deviation of DB Index assessed clustering effect. The findings tested the typical UCI Machine Learning Data Set and Data Set of high-dimensional industrial authenticity. The findings prove the efficiency of the new approach in a broad scale. The two parameters confirm the effect of the clustering.