Comparison of Tree Implementation, Regression Logistics, and Random Forest to Detect Iris Types

Abstract

Iris is a genus of 260-300 species of flowering plants with striking flower colors and has a dominant color in each region. The name iris is taken from the Greek word for rainbow, which is also the name for the Greek goddess of the rainbow, Iris. The number of types of iris plants with almost the same physical characteristics, especially in the pistil and crown, causes the misdetection of iris plant types. Iris plants are deliberately used because data is already available digitally on the internet and software such as orange and is widely used as a material for classifying objects. This research was conducted to classify iris plant types using three algorithms, namely Tree algorithm, Regression Logistics, and Random Forest. Classification algorithms are a learning method for predicting the value of a group of attributes in describing and distinguishing a class of data or concepts that aim to predict a class of objects whose class labels are unknown. The results showed the largest AUC (Area Under Curve) value obtained by the Random Forest method. AUC accuracy is said to be perfect when the AUC value reaches 1,000 and the accuracy is poor if the AUC value is below 0.500. As for the precision value of the three models used Random Forest has the highest precision value. From the data tests that have been done training and testing can be seen that the level of accuracy of testing of the three models where the Random Forest model is superior as a method for classification of irises.