网络口碑的跨平台分布与在线销售——基于BP人工神经网络的信息熵与网络意见领袖敏感性分析

The Distribution of Word-of-Mouth across Websites and Online Sales:Sensitivity Analysis on Entropy and Network Opinion Leaders Based on Artificial Neural Network

与以往通过评论数量、效价、离散度与中心性研究网络口碑对在线销售的影响不同,本文在充分考虑网络意见领袖影响力的基础上,引入了基于网络意见领袖的效价,从信息熵角度研究网络口碑的跨平台分布特征对在线销售的影响,并通过BP人工神经网络模型与现有研究成果进行了比较。结果发现,在多个网络平台并存且效价差异不大的情境下,传统效价不再是影响在线销售的重要因素,而基于网络意见领袖的效价、数量信息熵和效价信息熵才是影响在线销售的关键。进一步的BP反向传播网络模型分析结果显示,与现有理论相比,本文所构建的理论模型大大提升了预测的精确度。本文推进了网络口碑理论的发展进程,提供了网络口碑跨平台分布特征对在线销售影响的新证据,也为在线营销商产品投放与经营策略的制定提供了新的经验路径。

Empowered by the information technology and electronic commerce, modern consumers have available a vast of online vmouth （WOM） to inform their purchase decisions. Compared to traditional WOM, online WOM was perceived by consumers as more powerful, effective and unbiased. It also allows divergent opinions which come from different consumers presented on the website. Not only that, more and more consumers indicate a higher degree of trust on the online WOM. So many companies are allocating larger portions of marketing budgets to generate and manage the online WOM. However, the existing of different types of network platforms presents the online marketers with a dilemma. In order to benefit the companies most, how to choose the best network platform and manage the WOM become the focus issues to the managers. In order to solve these problems, most of extant re- search studies the attributes of online WOM hosted by one single site with mixed findings. The only exceptions in this line of research ignore the information of brand, price, involvement and influence of opinion leaders and also not solve the problems. For this, based on the previous achievements, with consideration of these factors, this pa- per exploits the relationships of the distribution of online WOM across websites and online sales. Compared to existing research, which studied the relationship between online WOM and online sales via vol- ume, valence, dispersion and centrality, from the perspective of entropy this paper introduces the valence based on the influence of network opinion leaders to research this question. By using data from jd. com, dangdang, corn, am- azon. cn and tmall, corn, this paper investigates the sales impacts of the distribution of WOM across websites. At the same time, in order to verify the accuracy and superiority of our model, used the B - P Artificial Neural Network, the article compares our theoretical model with the extant research. The main conclusions are as follows： firstly, in the context of little differences among multiple websites, valence which computed by the arithmetic average is no longer an important factor for online sales. In other words, the positive effect of favorable rate and the ＂negative bi- as＂ which have been verified by precious articles are not support in this context. Secondly, consistent with existing research, the volume of WOM have a positive effect on online sales. Thirdly, more evenly distributed online WOM across websites implies a higher search cost for WOM information and so has a negative impact on product online sales. But this effect is opposite for the entropy of valence. Specifically, a bigger entropy of valence implies more consistent users＇ reviews, so the search cost is lower compared to the smaller entropy of valence and it has a positive effect on online sales. Fourthly, according to the result computed by Back Propagation Neural Network model, the theory constructed by our paper is more accurate. In the end, this paper also controls the effect of brand, price and involvement of website. The results indicate that the impact of brand is limited in the network marketing context. Moreover, the contributions of our findings are as follows ： （ 1 ） based on the RFM model, the paper proposes a valence with the consideration of the network opinion leaders＇ influence. The empirical results show that this va- lence has a more accuracy than the valence that computed by arithmetic average. These findings contribute to broa- der research on online WOM. （2） With the consideration of the former factor, this study is the first to reveal that the distribution of online WOM across websites and gives same reasons about the contradiction in the extant research which always take one single website as a data source. Without a doubt, the results of latter is actually build on these conditions： the website that they used is based on a particular online WOM distribution. Different websites and even the same website in different times shows different distributions. So ignoring the distribution of WOM can＇ t get a consistent conclusion. This explanations provides a new perspective for existing research and also contribute to our understanding on the magnitudes of online WOM effects from multiple websites. （3） Based on the informa- tion theory, this paper introduces the entropy to study the attributions of volume and valence of WOM across web- sites, and studies the relationships of the distribution of WOM across websites and online sales. In the end, used Back Propagation Neural Network model, the empirical results show that the sensitivity and accuracy of our paper is better than existing researches. Finally, this paper not only affords some recommendations on the management of online Word-of-mouth for online marketers l especially on the condition of receiving positive feedback exclusively from one single website, but also provides some measures to promote a new product just entering the online market and a more mature product that has already received reviews on some websites.