Open AccessProduct review management software based on multiple classifiers
Author(s) -
Catal Cagatay,
Guldan Suat
Publication year - 2017
Publication title -
iet software
Language(s) - English
Resource type - Journals
ISSN - 1751-8814
DOI - 10.1049/iet-sen.2016.0137
Subject(s) - c4.5 algorithm , voting , computer science , majority rule , random forest , implementation , classifier (uml) , machine learning , decision tree , product (mathematics) , software , artificial intelligence , support vector machine , data mining , software engineering , naive bayes classifier , programming language , geometry , mathematics , politics , political science , law
In recent years, due to significant developments in online shopping and the widespread use of e‐commerce, competition among companies has increased considerably. As a result, product reviews have become a primary factor in consumers' decision making, which has given rise to a market for fraudulent reviews about real products and services. In this study, the authors propose a model using a multiple classifier system to identify deceptive negative customer reviews, which they validated with a dataset of hotel reviews from TripAdvisor. The proposed model used five classifiers by following the majority voting combination rule – namely, libLinear, libSVM, sequential minimal optimisation, random forest, and J48 – the first two of which represent different implementations of support vector machines. Ultimately, the model provided remarkable results that demonstrate improvement upon approaches reported in the literature.