Metode Smote untuk penanganan Imbalance Class pada Klasifikasi Data Tweet Pemilu Tahun 2024 di Indonesia menggunakan Algoritma Backpropagation

Abstract

Class imbalance is a classification problem where the target class distributions have very different ratios. The common thing that causes imbalance between classes is organic distribution (a type of distribution achieved through natural means). In this research, it was found that there was a class imbalance in the 2024 election tweet data in Indonesia, where there was more data with negative sentiment classes compared to positive and neutral sentiment. So a method is needed to deal with this problem, in this research the Synthetic Minority Oversampling Technique (SMOTE) method is used to overcome this problem. To test the accuracy of the model, two testing processes were carried out, first testing was carried out using the Backpropagation algorithm without considering class imbalance, and secondly the data was balanced first using SMOTE so that each class had the same amount, then tested again using the Backpropagation algorithm . The test results using Backpropagation obtained an accuracy value of 55% and a loss of 0.41, while tests carried out with SMOTE and Backpropagation obtained an accuracy value of 68% and a loss of 0.35. The application of SMOTE can overcome the problem of class imbalance and obtain better accuracy values.