Pete Andre Cadelina, Kashmel Galil Go, Glenn Pepito
The advent of the Internet has revolutionized the dissemination of information, leading to an abundance of both legitimate and false information. This information can shape public opinion, making it crucial to know whether that information is truthful or not. This study aims to build upon previous research by incorporating the latest OpenAI technology and traditional models. This study implemented traditional machine learning models, such as Multinomial Naive Bayes, Logistic Regression, and Support Vector Machine, with GPT 3.5 turbo fine-tuning. A comparison between the ensemble base model and ensemble base model with GPT3.5 was made. The researchers utilized the confusion matrix to evaluate and assess the performance of the models. The ensemble base model showed overall balanced metrics compared to the ensemble model with GPT. Furthermore, the ensemble base model showed an increase of 0.25% in accuracy as well as 0.46% increase in F1-score compared to the baseline. By ensembling traditional models, the study showed an increase of performance compared to the baseline model. © 2024 IEEE.
Information Science and Mathematics, University of San Carlos, Department of Computer, Cebu City, Philippines
