Performance Improvement of Paddy Growth Stage Classification Using Stacking CART and Random Forest Method

Indonesia is an agricultural country with one of its main agricultural commodities being rice. One of the areas that is the center of rice cultivation in Indonesia, especially on Java Island, is the Karawang Regency. There are several methods of measuring rice production data, one of which is the ASF statistical method. Apart from using these statistical methods, there are several studies showing that machine learning models can be developed to predict paddy growth stage precisely and accurately. This research uses Sentinel-1A image data which is extrapolated with ASF samples which are then oversampled as an imbalance data handling method. In this study, the development of the CART algorithm was carried out by using parameters maxnodes value of 1000 and a minleafpopulation of 1. Furthermore, the development of a stacking CART was carried out with the meta learner being a random forest. The results of this study indicate that when oversampling is applied to the data, Pearson Correlations in the estimated areas from the CART Method of classes V1 (Early Vegetative), V2 (Late Vegetative), G (Generative), and P&PL (Harvest and Land Preparation) toward estimated area derived from area sampling frame (ASF) method have reduced with deviation average is 0.21195 compared to Pearson Correlation in the estimated area using the same methods without oversampling the data. In addition, area estimation using the stacking CART+RF method shows an increase in Pearson correlation with an average of 0.26612 compared to the area estimation results using the CART method with oversampled data.