Spam review detection using self-organizing maps and convolutional neural networks

Online public reviews have significant influenced customers who purchase products or seek services. Fake are posted online to promote demote targeted reputation of the organizations and businesses. Spam review detection has been focus many researchers in recent years. As services growing rapidly, importance issue is ever increasing needs be addressed properly. In this regard, there a variety approaches that introduced distinguish truthful from fake ones. The main features engineered past studies typically involve two types linguistic-based behavioral-based characteristics reviews. Unsupervised, supervised semi-supervised machine learning methods widely utilized perform such classification. This paper introduces novel approach detect genuine ones using linguistic features. Unsupervised via self-organizing maps (SOM) conjunction with convolutional neural networks (CNN) employed classification We transform into images by arranging semantically-similar words around pixel image equivalently SOM grid cell. resulting consequently fed CNN for training then Comprehensive tests on gold-standard datasets show effectiveness proposed method single multi-domain contexts accuracy 88% 87%, respectively.