SCLAVOEM: hyper parameter optimization approach to predictive modelling of COVID-19 infodemic tweets using smote and...

Area of Science:

• Computational Social Science
• Artificial Intelligence
• Health Informatics

Background:

• The COVID-19 pandemic is accompanied by a dangerous infodemic of fake news and myths.
• Misinformation on social media, particularly Twitter, threatens public health efforts to control the pandemic.
• Effective detection of fake news is crucial for public safety and informed decision-making.

Purpose of the Study:

• To propose and evaluate a novel method for classifying fake COVID-19 tweets.
• To develop a hyperparameter optimization approach for predictive modeling of COVID-19 infodemic tweets.
• To enhance the detection of misinformation using the Synthetic Minority Over-Sampling Technique (SMOTE) and a classifier vote ensemble (SCLAVOEM).

Main Methods:

• Implementation of the Synthetic Minority Over-Sampling Technique (SMOTE) to address imbalanced datasets.
• Development of a classifier vote ensemble (SCLAVOEM) method for fake news classification.
• Application of hyperparameter optimization across model variables for predictive modeling.

Main Results:

• The SCLAVOEM model achieved weighted averages of 0.999 for F-measure and 1.000 for Area Under Curve (AUC).
• SMOTE integration led to performance increases ranging from 0.29% to 8.05% across three different datasets.
• The model successfully classified tweets into 'positive', 'negative', and 'click-trap' categories.

Conclusions:

• The proposed SCLAVOEM method provides an effective framework for the predictive detection of fake COVID-19 tweets.
• The integration of SMOTE significantly improves the performance of fake news classification models.
• This automated system can help mitigate the impact of misinformation and protect the public from information overload.