Analisis Optimasi Algoritma Decision Tree, Logistic Regression dan SVM Menggunakan Soft Voting

Abstract

Agriculture constitutes a fundamental pillar of a nation's economy. One key to success in agriculture is the selection of suitable land. The prediction of whether land is fertile or not can be efficiently accomplished through a data mining approach. This is because data mining offers several algorithms for extracting crucial information from vast datasets through classification. However, classification algorithms in data mining often encounter the challenge of data imbalance, which can lead to low accuracy rates. Processing data with calculation models that have low accuracy rates can result in numerous erroneous predictions (fail predictions). To address this issue, this research conducts testing and comparative analysis of the confusion matrix results from four calculation models: the Decision Tree algorithm, Logistic Regression, SVM, and the combination of these three algorithms using the Soft Voting ensemble technique. The test results indicate that processing data using the Decision Tree, Logistic Regression, and SVM algorithms, along with the optimization of the Soft Voting ensemble model, achieves the highest accuracy rate of 91.53%. This accuracy rate is higher compared to the other three calculation models: the Decision Tree algorithm with a difference of 3.83%, Logistic Regression with a difference of 2.66%, and SVM with a difference of 1.36%. This research makes a significant contribution by identifying an efficient solution to improve the accuracy of identifying fertile agricultural land, which is a crucial step in supporting the success of the agricultural sector in the country's economy.