Empirical Analysis And Modeling Of Online Reviews Evolutionary Pattern

Nowadays Online Consumer Reviews(OCR)has increasingly received scholars' attention as an important form of word-of-mouth.Recent study shows that online reviews of a product,such as a book or a restaurant,have effect on long-term consuming behavior and the future rating of the product,it mainly reflects that the early high rating of a product will lead the decrease trend of rating over time.To confirm the existence of the effect and explore how it works,over three million reviews on Dianping.com were collected to investigate the behavior patterns and intrinsic dynamics.In this paper,four temporal evolution patterns were observed via evaluating the cumulative average rating series for each restaurant.Moreover,a conceptual model considering the influence of consumer heterogeneity and the self-selection mechanism was introduced,and the numerical results coincide with the empirical analysis well enough to support the hypotheses.We find specific preferences result in tendentious consumption and unrepresentative reviews,which leads the potential consumers to over-or under-estimate the products and directly affect the subsequent ratings.The results of the model showed that the cumulative average rating series tend to be stable,but in fact,the cumulative average ratings of the restaurants tend to fluctuate after the trend of being stable.There are many reasons for the fluctuations,self-boosting rating manipulation is one of them,and it means some merchants improve their ratings to attract more consumers by mix a large number of fake high rating reviews into the real consumer reviews in the word-of-mouth platform.Self-boosting rating manipulation was introduced into our model to observe the influence of rating manipulation on the evolution of cumulative average rating.The results showed that although the cumulative average rating would rapidly increase in a short period due to the influence of self-selection mechanism and individual preference heterogeneity,it would soon fall back,so the impact of self-boosting rating manipulation is not permanent.The conclusions of this paper can contribute to the specific policies to adjust the initial rating effect for the specific marketing strategies,and can provide ideas of the rating manipulation recognition.