Using and comparison of artificial intelligence techniques to detect misinformation and disinformation on Twitter

This study addresses a widespread problem of the potential impact of misinformation and fake news on public discourse by examining a set of artificial intelligence (AI) techniques to detect misinformation on the social media platform Twitter. A variety of models, including Long Short-Term Memory (LSTM), Support Vector Machine (SVM), Random Forest (RF), Gaussian Naive Bayes (GaussianNB), Gradient Search (GBoost), Decision Tree (DT), Logistic Regression (LR), Extreme Gradient Search (XGBoost), and K-Nearest Neighbors (K-NN), are used to distinguish deceptive and trustworthy content. The analysis uses natural language processing (NLP), deep learning (DL), and machine learning (ML) techniques to demonstrate the effectiveness of each model in detecting misinformation patterns using a dataset consisting of 23481 fake tweets and approximately 21418 genuine tweets. The study provides a comprehensive assessment of the strengths and limitations of these AI systems, particularly focusing on accuracy, efficiency, and scalability. The results show that XGBoost performs best with 99.82% accuracy and 99.81% F1-score. This is followed by (GBoost) with 99.63% accuracy and 99.62% F1-score and (DT) with 99.61% accuracy and 99.59% F1-score. The results show that these models are the most effective models. While the other models show accuracy ranging from 99.31% to 81.63%, the results provide insight into the performance of the main models, contributing significantly to the fight against disinformation and the reliability of information.