LDCP: A Novel Approach to Predict Fake Reviews in Online Social Network by using Learning based Data Classification Principle

The proliferation of online reviews has significantly impacted consumer decision-making processes, making the detection of fake reviews crucial for maintaining trust in e-commerce platforms. This study presents a novel approach for detecting fake reviews using machine learning (ML) and deep learning (DL) models, with a focus on achieving high accuracy through advanced feature extraction techniques and data augmentation. Various models, including classical ML algorithms such as Support Vector Machine (SVM) and Random Forest, alongside DL models like Convolutional Neural Networks (CNN), Bidirectional Long Short-Term Memory (BiLSTM), and Cascaded Multi-Layer Perceptron (MLP), were evaluated. Feature extraction methods like TF-IDF, GloVe, and combined approaches were used to enhance performance. The DL model RoBERTa achieved the highest accuracy of 99.5%, while Cascaded MLP reached 99.2%. Classical models like Random Forest and SVM also performed well, achieving accuracies of 97.1 % and 96.8%, respectively. Data augmentation techniques, such as back-translation, further improved the models' ability to handle imbalanced datasets. The results demonstrate that combining DL architectures with advanced feature extraction leads to superior performance in fake review detection.