Kimberlites Identification by Classification Methods

Kimberlites identification is a very important task for diamond mining. In traditional way, geologists draw upon past experience to do this work. Whether the bedrock should be drilled depends on their analysis of rock samples. This method has two disadvantages. First, as the database increasing, it becomes more difficult to do this work by manual inspection. Secondly, the accuracy is influenced by the expert’s experience, and it reaches scarcely 80 percents averagely. So an analytical method to kimberlites identification over large geochemical datasets is demanded. This article applies two methods (SVM and decision tree) to a dataset provided by a mining company. Comparing the performances of these two methods, our results demonstrate that SVM is an effective method for this work.