An Approach to Improve the Accuracy of Detecting Spam in Online Reviews

Abstract

Customers or user opinion is the most important and valuable information at online nowadays, especially in product reviews. Mostly customer used to make their decision for purchasing a particular product based on the other's customer reviews. Those reviews are increasing the rating of that e-commerce site. Normally, reviews are considered unbiased opinion of a person who has personal experience with a related specific product. The noticeable thing is that many reviewers reviews are not real or authentic. These kinds of feedback are usually called spam, and it is becoming a large problem in online and other electronic communication. As the value of online reviews is getting increased, the spammers are getting inspired to doing spam for them or promoting a specific e-commerce website. Also, they are demoting a specific site for payment. In this paper, we have discussed some traditional techniques for detecting spam in online public opinions. Next, we have used the stacking algorithm with some traditional classifiers for the detection of spam reviews. Finally, the performance of different classifiers has evaluated through a simulation experiment. From the experiments, we have seen that stacking classifier provides better accuracy than other traditional classifiers.