Detecting fake reviews is essential for maintaining the credibility of e-commerce platforms and protecting consumers from misleading information. However, this study specifically contributes by providing a comparative analysis of the EKI-SM and BSTC models, highlighting the efficiency and learning stability advantages of the EKI-SM in fake review detection tasks. Using a dataset of 40,000 Amazon product reviews, we conducted experiments to evaluate both models based on accuracy, precision, recall, F1-score, and AUC score. The results show that EKI-SM achieves an accuracy of 94.98%, recall of 91.82%, and F1-score of 94.85%, slightly outperforming the BSTC, which achieves an accuracy of 94.60 %. Although the BSTC showed marginally higher precision, the difference was not significant. Beyond performance metrics, this study emphasizes the conceptual contribution of EKI-SM, which provides a better balance between precision and recall, as well as faster convergence and lower training loss. Both models achieved high AUC scores (99.41% for BSTC and 99.42% for EKI-SM), confirming their strong capability to distinguish between genuine and fake reviews. These findings indicate that EKI-SM is not only competitive in classification performance but also more efficient and stable during training, making it a more reliable approach for fake review detection