COMPARATIVE ANALYSIS OF AUTHORSHIP ATTRIBUTION USING SUPPORT VECTOR MACHINE AND NAIVE BAYES

Abstract

This study investigates authorship attribution using Support Vector Machine (SVM) and Naive Bayes algorithms on the “indosaya.csv” dataset within the CRISP-DM framework. The dataset contains 219 Indonesian text entries categorized into four author labels: “Dia,” “Ayah,” “Saya,” and “Paman.” Data preprocessing was conducted using TF-IDF weighting and Sastrawi stemming techniques.

The results show that SVM achieved higher accuracy and better robustness in identifying writing style patterns compared to Naive Bayes. Although Naive Bayes offered faster computation, it was less effective in capturing relationships between words that characterize individual writing styles. Overall, the study demonstrates that SVM is more reliable for authorship attribution and forensic linguistic analysis in Indonesian text classification tasks.