Classification of Paddy Growth Phase with Machine Learning Algorithms to Handle Imbalanced Multi-Class Big Data

Abstract

The global Sustainable Development Goals (SDGs) adopted by countries in the world have significant implications for national development planning in Indonesia in the period 2015 to 2030. The Agricultural sector is one of the most important sectors in the world and has a very important contribution to achieving the goals. Availability of accurate paddy production data must be available to measure the level of food security. This can be done by monitoring the growth phase of paddy and predicting the classification of its growth phase accurately and precisely. The paddy growth phase has 6 classes with the number of class members usually not the same (imbalanced data). This study describes the results of the classification of paddy growth phases with imbalanced data in Bojonegoro Regency, East Java in 2019 using machine learning algorithms on the Google Earth Engine (GEE) platform. Classification is done by Classification and Regression Tree, Support Vector Machine, and Random Forest. Oversampling technique is used to deal the problem of imbalanced data. The Area Sampling Frame survey in 2019 conducted by BPS was used as a label for classification model training. The results showed that the overall accuracy (OA) using the Random Forest algorithm by modifying the dataset using oversampling was 82.30% and the kappa statistic was 0.76, outperforming the SVM and CART algorithms.