Fake Review Detection Based On Ensemble Learning

With the rapid development of Web 2.0 technology in the Internet era,network users can create their own comment content,which contains a wealth of information about network products.Unfortunately,due to personal interests,some businesses or individuals post deceptive(fake reviews)reviews to glorify themselves or disparage their competitors.These behaviors also mislead potential customers about their spending habits.Due to the strong domain relevance of online comments,different comment fields not only comment on texts but also comment on behaviors.Therefore,different detection methods must be used to detect fake comments in different fields.This paper divides online reviews into product review and store review.Since there are few detection methods for false store review,it is an important research topic to analyze the characteristics of store fake review and design effective detection methods.In this paper,we propose an integrated learning-based approach to store fake review detection.There are three stages of this approach.In the first stage,we focus on store fake review detection problem by exploiting the labeled datasets containing hotel and restaurant reviews from Yelp.com.We divide the feature of store review into two categories,behavioral feature and content feature,and then analyzes the validity of these feature.In the second stage,the data sets are sampled by cross validation and Under-sampling respectively.In the third stage,the Random Forest and LightBGM algorithms in ensemble learning were used to perform supervised fake review detection on their comments respectively with the Decision Tree,KNeighbors and naive bayesian based on gaussian distribution in traditional machine learning algorithm,and the two methods were compared.Balance the detection precision and recall ra by adjusting the parametersThe experimental results show that the ensemble learning algorithm has higher precision than the traditional machine learning algorithm,which is of great significance for the detection of fake store review.