<![CDATA[The development of information technology has changed the way people interact in the digital space, including through the Twitter platform, which is widely used to share information and opinions. However, this convenience has also led to the emergence of fraudulent content such as fake investments, fictitious sweepstakes, and fictitious donation requests. This study aims to analyze and compare the performance of five machine learning algorithms, namely Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), and Naïve Bayes (NB), in detecting fraudulent Indonesian language content on Twitter. The dataset consists of 5.221 Indonesian language tweets that have been manually labeled into two classes, fraud and non-fraud. All tweets were processed through text preprocessing stages, including data cleaning, case folding, normalization, tokenization, filtering, and stemming, before being represented as numerical vectors using Word2Vec. Classification was performed using 10-fold cross-validation with evaluation metrics of accuracy, precision, recall, and F1-score. The results show that Random Forest achieved the best performance with accuracy of 85.6%, followed by SVM (84.0%), Logistic Regression (83.6%), Decision Tree (81.2%), and Naïve Bayes (78.4%). The main contribution of this study is to provide a systematic empirical comparative analysis of classification algorithms for detecting Indonesian fraudulent content on Twitter, which remains relatively underexplored. These findings show that the combination of Word2Vec and Random Forest can effectively capture the semantic context of short texts and can serve as a reference for developing automatic detection systems for fraudulent content on social media.]]>
Copyrights © 2026
