Identification of Java Tea Adulteration by Babadotan and Tekelan using Machine Learning

Abstract

Java Tea (Orthosiphon aristatus) is a common herbal medicinal plant that functions as a health treatment and treats various diseases. The high demand for Java Tea causes high prices and a decrease in the amount of medicinal plant raw materials, causing various quality control problems such as the content of various bioactive components and adulteration from babadotan and tekelan. So far, the detection of adulteration has been carried out by various analyzes, including chemical analysis and statistical methods to process data. The data used is of high dimension with a very high-density level, thus causing difficulties in classification. The mixed data of Orthosiphon aristatus consists of 1201 features with a total sample of 216. This study uses a Random Forest (RF) method with a total of 100 trees, and the RF method is combined with the Recursive Feature Elimination (RFE) method. In the RF and RFE that can be produced, the optimum value for the number of features is 244. The experimental evaluation results revealed that the proposed method could achieve a high accuracy of 81.82% in identifying Orthosiphon aristatus.