Research On Reliability Prediction Of Online Product Reviews Based On DDAG-SVM

As an important electronic word of mouth, Online product reviews (OPR) can significantly reduce asymmetric information between customers and online merchants, thus playing a vital role on customer shopping decision online. However, in the era of Big Data, there are millions of top OPR on the online shopping platform and their quality varies a lot, so customers are faced with great challenges caused by the information overloading problem of OPR to learn more about what they want to buy online. Based on analyzing the influence factors of reviews reliability, this paper proposes the Support Vector Machine Based Decision Directed Acyclic Graph (DDAG-SVM) model to realize the reliability prediction of OPR.Firstly, this paper introduces literature reviews about the existing relevant research on OPR from several aspects, namely OPR incentives, OPR influence on customers shopping decision, research on reviews reliability, reviews opinion mining and text classification. Then, this paper comprehensively describes the four formation path body and six formation paths in the formation process of fake OPR and obtains the main features of different types of fake reviews. Based on the analysis above, ten-influence-factor index system has been built from reviews content, reviewer features and merchant features. Review attribute of product-based emotion evaluate is the main innovation points of this paper. Next, a DDAG-SVM model is proposed to predict reliability of OPR based on 10 feature vectors. Finally, based on MATLAB and LIBSVM, this paper conducts the contrast experiments by approximately 5000 reviews crawled from Taobao website, and the correct rate of this model is 93.687%.The experiment result shows that it has higher feasibility and correct rate.The research on reliability prediction of OPR in this paper, significantly decrease information processing ability demand for customers to make right shopping decision, Additionally, it also helps e-commerce platform improve their review system. Therefore, the results in this research have great theoretical and practical significance on improving OPR system and promoting online shopping experience for customers.