Detection of fake reviewer groups in product reviews using novel bidirectional encoder revolutionary transformer compared with stochastic gradient descent Available to Purchase

The objective of this study is to develop an effective model for detecting extremist reviewer groups online, employing a Novel Bidirectional Encoder Revolutionary Transformer (BERT) in comparison to Stochastic Gradient Descent. Both algorithms are applied with a fixed number of iterations set to 20 for each group, aiming to enhance the accuracy of predicting the online reviewer system. The dataset used for evaluation is Amazon Review, responsible for determining the sentiment of product reviews (positive or negative). It encompasses reviews from various categories: books (2834 samples), DVDs (1199 samples), electronics (1883 samples), and kitchen and housewares (K). The performance evaluation metric is accuracy, with a desired statistical power (G power) of 0.8 and a confidence level of 95%. Statistical analysis reveals a significant difference between the two groups, with the Novel Bidirectional Encoder achieving an accuracy of 96.75%, outperforming Stochastic Gradient Descent, which attains 85.05% accuracy. This difference is statistically significant, with a significance value of 0.01 (p<0.05). Consequently, the accuracy percentage of the Novel Bidirectional Encoder Revolutionary Transformer surpasses that of Stochastic Gradient Descent.