Identification of Fake Comments in E-Commerce Based on Triplet Convolutional Twin Network and CatBoost Model

The continuous progress of Internet technology has promoted the maturity of online shopping market. The false comments accompanying online shopping not only infringe the rights and interests of consumers, but also pose a threat to the healthy development of e-commerce. To promote the development of e-commerce, a false comment recognition model based on triplet convolutional twin network and CatBoost model is proposed. Key semantic features are refined from the perspective of semantic similarity through triplet convolutional twin networks. Finally, integrating the advantages of the CatBoost model and combining text and behavioral data aims to reduce over-fitting and enhance recognition accuracy. It provides an efficient and accurate identification tool for e-commerce platforms. The benchmark experimental results show that the proposed TriCNN-CatBoost model significantly outperforms traditional Naive Bayes, Support Vector Machines, and Random Forest models in terms of accuracy, recall, and F1 score, demonstrating stronger false comment recognition ability and generalization performance. In addition, the proposed model not only converged quickly, but also had the lowest loss values on two different datasets. The F1 scores on the four key features were 0.8136, 0.8267, 0.8046, and 0.7966, all of which were superior to other comparison models. At the same time, the accuracy of the four features was 0.8931, 0.9012, 0.8846, and 0.8961, respectively, which verified the excellent predictive recognition performance. Compared with the other two comparison algorithms, the detection time required by the proposed model was the shortest at 115us and consumed the least amount of resources. Overall, the model has improved the recognition performance, proving its effectiveness in providing consumers with a more accurate and reliable shopping experience, while also providing platform managers with a scientific and efficient regulatory tool.