Optimalisasi Klasifikasi Support Vector Machine dengan SMOTE: Studi Kasus Ulasan Pengguna Aplikasi Alfagift

Support Vector Machine (SVM) is a supervised learning algorithm that works by classifying based on classes that refer to patterns resulting from the training process. SVM has several commonly and popularly used kernels, one of which is the linear kernel. The weakness of SVM is in the "parameter selection" and its performance tends to be poor in the case of unbalanced datasets. The purpose of this study is to overcome the weaknesses of the SVM algorithm with the proposed method. This research uses a linear kernel with feature extraction that is Word2Vec with Skip-gram model, and in handling the data imbalance problem using SMOTE (oversampling) technique. The results showed that the unbalanced dataset produced an accuracy of 90% and the balanced dataset (SMOTE) produced an accuracy of 92%, so the SMOTE oversampling technique was proven to increase the accuracy results by 2%.