Classification Comparison Performance of Supervised Machine Learning Random Forest and Decision Tree Algorithms Using Confusion Matrix

Abstract

The classification method is part of data mining which is used to predict existing problems and also as predictions for the future. The form of dataset used in the classification method is supervised data. The random forest classification method is processed by forming several decision trees and then combining them to get better and more precise predictions. while a decision tree is the concept of changing a pile of data into a decision tree that presents the rules of a decision. From these two classification methods, researchers will compare the level of accuracy of predictions from both methods with the same dataset, namely the employee dataset in India, to predict the level of accuracy of employees who leave their jobs or still remain to work at their company. The number of records available is 4654 records. Of the existing data, 90% was used as training data and 10% was used as test data. From the results of testing this method, it was found that the accuracy level of the random forest method was 86.45%, while the decision tree method was 84.30% accuracy level. Then, by using the confusion matrix, you can see the magnitude of the distribution of experimental validity visually to calculate precision, recall and F1-Score. The random forest algorithm obtained precision of: 96.7%, sensitivity of: 84.7%, specificity of: 91.4%, and F1-Score of: 90.2%. Meanwhile, the decision tree algorithm obtained precision of: 95.7%, sensitivity of: 82.9%, specificity of: 88.4%, and F1-Score of: 88.8%.