Modelling a dense hybrid network model for fake review analysis using learning approaches

• Published in: Soft computing (Berlin, Germany) Germany), 2024-02, Vol.28 (4), p.3519-3532
• Main Authors: Srisaila, A., Rajani, D., Madhavi, M. V. D. N. S., Asha Shiny, X. S., Amarendra, K.
• Format: Article
• Language: English
• Subjects: Application of Soft Computing; Artificial Intelligence; Computational Intelligence; Control; Coronaviruses; COVID-19; Engineering; False information; Machine learning; Mathematical Logic and Foundations; Mechatronics; Performance assessment; Robotics; Social networks
• ISSN: 1432-7643; 1433-7479
• DOI: 10.1007/s00500-023-09609-4

Users are now able to provide opinions and feedback in the form of reviews for any product, service, or business on social networking and e-commerce websites. Due to the significant user effect of reviews, spammers utilize phony reviews to elevate their organization or product and denigrate their rivals. On any given platform, it is thought that 14% of the reviews are fraudulent. To identify bogus reviews, several academics have put forth several strategies. The drawback of existing techniques is that they analyze the entire review text, which lengthens calculation times and reduces accuracy. In our suggested method, just these elements and their corresponding feelings are used for the detection of phony reviews. Aspects that have been retrieved are sent to CNN for learning. To detect false reviews, the reproduced attributes are input into long short-term memory (LSTM). As far as we are aware, despite the optimization it provides, aspects of replication and extraction are not employed to detect fake reviews, which is a big contribution from us. Performance comparisons with more modern methods are done using the Ott and Yelp Filter datasets. Analysis of the results of experiments shows that our suggested strategy beats current strategies. To demonstrate that dense hybrid network models (d-HNM) are superior to established machine learning techniques for difficult computing problems, our approach is also contrasted with others.