Abstract—In this paper we have presented a new method to enhance the clustering algorithm. Clustering is a method to label points in a two or more dimensional scatter plot. Some algorithms such as FCM, KNN or K-means implement this, but the problem occurs in the case that the clusters are close to each other or even they have overlaps (like neural data which we first encountered). In this state the human observer can detect some clusters on the scatter plot but the algorithms fail or they offer different clusters. To solve this problem we propose that the TPS transform on the primary data (clusters) can improve the clustering solution which is more likely to human perception of the clusters. To show the validity of our method we have used the SPSS analyses.