Abstract: The abundant data resources embedded in e-commerce platforms provide consumers with valuable references for decision-making and constitute an important factor influencing purchasing behavior. The effective use of product-related data resources can help online retailers optimize pricing strategies. From the perspective of consumer social learning, this paper builds on expected utility theory and considers how consumers enhance product cognition by sequentially browsing online reviews during the purchase decision process. The cognitive updating process is quantified using the anchoring effect and Bayesian theorem, based on which a product revenue-based pricing model is developed. In addition, deep learning methods and Python-based web crawling techniques are employed to conduct four-category sentiment analysis of review texts. This paper examines the impact of the ordering of positive and negative review information and consumers’ “decision threshold” on customer cognition, product pricing, sales volume, and profits. The results demonstrate that, contrary to prevailing conventional wisdom, compared with products with no reviews, a small, moderate, and intermittently appearing amount of negative information can actually be more beneficial for product pricing and marketing. This research provides a theoretical foundation and decision support for data-driven product pricing and operational management by online retailers.

Key words: online product pricing, social learning, anchoring effect, deep learning, Bayesian theorem

摘要： 电商平台蕴藏的丰富数据资源可为消费者提供决策参考，是影响消费者购买行为的重要因素；同时，合理利用产品数据资源也有助于在线零售商优化定价策略。本文从消费者社会学习视角出发，基于期望效用理论，考虑消费者在购买决策中通过逐条浏览在线评论来加强产品认知，运用锚定效应与贝叶斯规则量化认知更新过程，构建产品收益定价模型。借助深度学习方法和 Python爬虫技术，对评论文本进行四分类情感分析，研究正负评论信息排列顺序与消费者“决策临界值”对消费者认知、产品定价、销量与利润的影响。研究结果表明：不同于传统观点，相较于零评价，少量、适度且不连续出现的负面信息反而有利于产品的定价与营销。本研究可为在线零售商基于数据驱动的产品定价决策与运营管理提供理论基础与决策支持。

关键词: 在线产品定价, 社会学习, 锚定效应, 深度学习, 贝叶斯规则