Prediction of Tobacco Leave Grades with Ensemble Machine Learning Methods

For many years, the Indonesian economy is influenced by the role of tobacco. It is not only for international trade but also for the farmers who plant the tobacco. However, to find a good tobacco grade is not easy. Many factors affect tobacco leaves grade. This paper focuses on developing a machine learning method to predict and determine the tobacco grade based on the environment condition and the plantation. Four independent variables that are temperature, sunlight hours, humidity, rainfall, and the plantation were used as a predictor to one target variable, which is the tobacco leaves grade. We applied two regression methods: Random Forest and Gradient Boosting Machine to predict whether there is a relationship between independent and dependent variables. The results depicted that Gradient Boosting Machine and Random Forest methods could be done to predict the tobacco grade successfully. The result also showed that Gradient Boosting Machine is superior to Random Forest in two experiments (with and without the plantation variables). Finally, to find the influenced variable for predicting the tobacco grade, i.e. sunlight hours has been performed.