Prediction Model for Soybean Land Suitability Using C5.0 Algorithm

Abstract

Soybean is one of the protein main sources that can be used for consumption in tempeh, tofu, milk, etc. Based on projection results, soybean production and consumption balance in Indonesia, in 2018-2022, it is estimated that deficit will increase by 6.18% per year. So, it's necessary to guide soybean land suitability, which can be carried out by evaluating existing land suitability to support soybean farming expansion and production. This study conducted an analytical study to evaluate soybean land suitability using C5.0 algorithm based on land and weather characteristics. The C5.0 algorithm is an extension of spatial decision tree, an ID3 decision tree extension. Dataset is divided into two categories: explanatory factors representing seven land characteristics (drainage, land slope, base saturation, cation exchange capacity, soil texture, soil pH, and soil mineral depth) and two weather data (rainfall and temperature), and a target class represent soybean land suitability in two study areas, namely Bogor and Grobogan Regency. The result generated two land suitability models with the best model obtained accuracy for training data 98.58%, while testing data was 97.17%. The best model rules are 69 rules that do not involve three attributes: cation exchange capacity, soil mineral depth, and rainfall.