Abstract: Due to the efficiency of data mining science for analyzing, reviewing extensive data, especiallygeochemical data, essential methods ,techniques such as the hierarchical method, K-Means method, densitybased methods, Cohennon method, so forth, have been developed ,utilized by numerous researchers forclustering. One of the most notable ,widely used algorithms in the field of clustering is the K-Means algorithm.This algorithm divides the data in K clusters by emphasizing the distance criterion. This study focuses on applyingthis method according to lithogeochemical data taken from the 1:100,000 scale map of Chahar-Farsakh in SouthKhorasan province for the elements of copper, cobalt, nickel to the sampling coordinates. The optimal value ofK was classified according to the desirability of the selection, the data, thus the relationships between theseelements in the range were determined. This was analyzed by changing the value of K from 3 to 15 criteria mentionedin each class to reveal the optimal K. According to the observations, the existence of a quadratic relationship withnegative concavity between copper, cobalt elements, as well as a special exponential relationship between copper, nickel, a positive linear relationship between nickel, cobalt, were reported. Finally, considering thecoordinates of the samples, the concentration of cobalt, nickel, the quantity of copper was predicted usinga General Regression Neural Network (GRNN). The accuracy of this method was estimated to be 0.99 on trainingdata 0.76 on test data. Therefore, using the proposed method (K-means Clustering, GRNN) in this paper, itis possible to examine the extent of changes in other elements in the analysis. Also, it is possible to make deeper,broader explorations via determining the relationship between the elements.