Product2Vec

通过机器学习理解产品竞争

主讲人

陈芳琳

讲座时间

2023年4月21日（周五）10:00

讲座方式

Zoom：968 1693 1357

主讲人介绍

陈芳琳于纽约大学获得市场营销学位博士和硕士学位，于清华大学获得经济学和数学学士学位，现为迈阿密大学市场营销系助理教授。她的研究聚焦于市场营销和机器学习的交叉领域，例如产品竞争和消费者定位。在她的一组研究中，陈芳琳博士专注于从产品共同购买模式中发掘信息，以理解产品关系和市场结构。她的另一组研究关注的是如何通过机器学习优化锁定目标顾客的顺序，从而提高客户留存率。此外，她还对媒体消费感兴趣，包括传统媒体和数字媒体。

Fanglin Chen is an Assistant Professor of Marketing at the University of Miami. Her main research interests center around the intersection of marketing and machine learning, specifically in the areas of product competition and consumer targeting. In one research stream, she focuses on extracting information from product co-purchase patterns to understand product relationships and market structure. The other research stream focuses on developing sequential targeting strategies to boost customer retention via machine learning. She is also interested in studying media consumption, including both traditional media and digital media. She received her PhD and MS in Marketing from New York University and BS in Economics and Mathematics from Tsinghua University.

讲座内容

从产品层面而非品牌层面出发去研究竞争和市场结构可以帮助企业更好的理解同类相食和产品线优化等问题。本研究引入的基于表征学习的Product2Vec方法适用于研究产品数量较大时的产品级竞争。该模型以购物篮为输入，为每个产品生成一个保留了重要产品信息的低维向量。基于对这些产品向量的分析，本研究得到了如下发现。首先，本研究证明这些向量可以反映任意一对产品之间的类比关系。其次，本研究创建的两个指标，互补性和互换性，能够准确的衡量任意一对产品之间的互补/替代关系。第三，本研究还通过将价格因素排除在产品向量之外调整了表征学习算法，从而进行价格弹性的估计并研究产品层面的竞争。研究表明，与普通的选择模型相比，本方法可以更快更准确的生成需求预测和价格弹性。第四，本研究提出了Product2Vec在营销实践上的两个应用：1）分析品牌内部和品牌间竞争；2）分析市场结构。总的来说，本研究展示了机器学习算法（例如：表征学习）在增强和改进传统营销方法方面的价值。

Studying competition and market structure at the product level instead of brand level can provide firms with insights on cannibalization and product line optimization. We introduce Product2Vec, a method based on representation learning, to study product-level competition when the number of products is large. The proposed model takes shopping baskets as inputs and, for every product, generates a low-dimensional vector that preserves important product information. Using these product vectors, we present several findings. First, we show that these vectors can recover analogies between product pairs. Second, we create two measures, complementarity and exchangeability, that allow us to determine whether product pairs are complements or substitutes. Third, we combine these vectors with traditional choice models to study product-level competition. To accurately estimate price elasticities, we modify the representation learning algorithm to remove the influence of price from the product vectors. We show that, compared with state-of-the-art choice models, our approach is faster and can produce more accurate demand forecasts and price elasticities. Fourth, we present two applications of Product2Vec to marketing problems: 1) analyzing intra- and inter-brand competition and 2) analyzing market structure. Overall, our results demonstrate that machine learning algorithms, such as representation learning, can be useful tools to augment and improve traditional marketing methods.