PERBANDINGAN PERFORMA SUPPORT VECTOR MACHINE DAN NAÏVE BAYES CLASSIFIER DALAM KLASIFIKASI EMOSI ULASAN PRODUK E-COMMERCE

THE COMPARISON OF SUPPORT VECTOR MACHINE (SVM) AND NAIVE BAYES CLASSIFIER (NB) ON CLASSIFIYING EMOTIONAL REVIEW ON E -SCOMMERCE

Ulasan yang ada pada e-commerce dapat mempengaruhi keputusan pembeli dan menunjukkan emosi pembeli terhadap produk yang mereka beli. Ulasan tersebut dapat diolah oleh produsen e-commerce untuk mengetahui kesan pembeli terhadap produk mereka. Namun, jumlah pengguna dalam e-commerce mencegah produsen untuk mengolah ulasan tersebut secara manual. Maka dari itu, dilakukanlah penelitian dengan Support Vector Machine (SVM) dan Naïve Bayes Classifier (NB) untuk meninjau ulasan-ulasan yang ditinggalkan para pembeli serta membandingkan kedua algoritma dalam melakukan klasifikasi. Dari hasil penelitian tersebut, secara keseluruhan ditemukan bahwa accuracy tertinggi diraih oleh SVM kernel Linear dan Sigmoid, yaitu 66,30%, precision tertinggi oleh SVM kernel rbf, yaitu 67,25%, recall tertinggi oleh SVM kernel Sigmoid, yaitu 62,12%, dan f1-score tertinggi oleh SVM kernel Sigmoid, yaitu 62,91%. Sedangkan NB dapat meraih accuracy tertinggi yaitu 63,33% oleh NB Multinomial dengan data resampling, precision tertinggi yaitu 66,97% oleh NB Multinomial, recall tertinggi yaitu 61,91% oleh NB Multinomial dengan data resampling, dan f1-score tertinggi yaitu 61,65% oleh NB Multinomial dengan data resampling. Selisih nilai tertinggi accuracy SVM dengan NB adalah 2,97%, sedangkan selisih nilai tertinggi precision SVM dengan NB adalah 0,28%, selisih nilai tertinggi recall SVM dengan NB adalah 0,21%, dan selisih nilai tertinggi f1score SVM dengan NB adalah 1,29%.

Reviews in e-commerce can affect buyers’ decisions and show their emotions toward the products they buy. E-commerce companies can process the reviews to the buyer’s impressions of their products. Unfortunately, the number of e-commerce users prevents companies from manually processing those reviews. Therefore, research is done using a Support Vector Machine (SVM) and Naïve Bayes Classifier (NB) to review the reviews left by the buyers and compare the two algorithms in doing classification tasks. The study revealed that the highest accuracy score was achieved by SVM using linear and sigmoid kernel, which was 66.3%, the highest precision score was reached by SVM rbf kernel, which was 67.25%, then the highest recall score was reached by SVM sigmoid kernel, which was 62.12%, and the highest f1-score was reached by SVM sigmoid kernel which was 62.91%. On the other hand, the highest accuracy was reached by Multinomial NB with resampled data with a score of 63.33% the highest precision was reached by Multinomial NB with a score of 66.97%, the highest recall was reached by Multinomial NB with resampled data with a score of 61.91%, and the highest f1-score was reached by NB Multinomial with resampled data. The difference between the highest accuracy value of SVM and NB is 2,97%, while the difference between the highest precision value of SVM and NB is 0.28%, while the difference between the highest recall value of SVM and NB is 0.21%, and the difference between the highest values of f1-score is 1.29%