EHT-DL: An Efficient Hyperparameter-Tuned Deep Learning Model for Fake News Detection

Abstract

Fake news detection is a critical task in today's information age, as the spread of misinformation can have severe consequences on individuals, societies, and even global events. The necessity of fake news detection arises from the growing challenges posed by the proliferation of false information across various platforms. Existing fake news detection techniques, such as Rule-Based Systems, Machine Learning Approaches, and Social Network Analysis, have made significant progress but still suffer from limitations in accurately capturing complex patterns and differentiating between real and fake news. This paper aims to address these limitations by proposing an efficient hyperparameter-tuned deep learning model (EHT-DL) for fake news detection. The EHT-DL model leverages a multi-step approach to effectively detect fake news. It begins with preprocessing steps such as text normalization, special character handling, tokenization, stop word removal, stemming, and lemmatization. This ensures the dataset is clean and ready for subsequent processing. Feature extraction is performed using word embeddings, N-grams, and TF-IDF scores to capture semantic information and word importance. The dataset is then split into training and testing sets, and the Dl4jMlpClassifier deep learning model is employed for classification. To tackle the drawbacks of existing techniques, the EHT-DL model incorporates efficient hyperparameter tuning. It applies both Grid Search and Random Search techniques to optimize the hyperparameters of the Dl4jMlpClassifier. By iteratively exploring various combinations of hyperparameters, the model identifies the best options that yield superior performance. This approach improves the model's accuracy and enhances its ability to differentiate between real and fake news. Experimental results demonstrate the efficacy of the EHT-DL model. The model is evaluated using standard evaluation metrics such as accuracy, precision, recall, and F1-score. Comparisons with existing techniques highlight the superiority of the proposed model in detecting fake news accurately and efficiently (83.27 % accuracy, 80.62 % precision, 71.57 % recall, and 75.83 % f1-score). The EHT-DL model achieves significant improvements in terms of performance, demonstrating its effectiveness in combating the challenges of fake news detection.