Optimization Accuracy of Fake News Detection for Twitter Data using Deep Learning Approach

Abstract

With the explosive growth of social media platforms like Twitter, the spread of fake news and misinformation has become a critical concern for digital trust and online safety. Detecting such misinformation accurately and efficiently requires advanced computational intelligence techniques. This study focuses on improving the optimization accuracy of fake news detection using deep learning models trained on Twitter data. Various neural architectures—such as Long Short-Term Memory (LSTM), Bidirectional LSTM (BiLSTM), Convolutional Neural Network (CNN), and hybrid CNN-BiLSTM models—are explored for text-based fake news classification. The dataset undergoes preprocessing steps including tokenization, stop-word removal, and feature embedding through Word2Vec and GloVe representations. Hyperparameter optimization techniques such as Grid Search and Bayesian Optimization are implemented to enhance model performance. Evaluation metrics including accuracy, precision, recall, F1-score, and ROC-AUC are used to measure performance. The results reveal that optimized hybrid models achieve significantly higher accuracy and generalization ability compared to traditional classifiers. The study concludes that combining deep learning and optimization techniques provides a robust, scalable, and interpretable framework for reliable fake news detection on Twitter.