Predictive Analytics Of E-Commerce Search Behavior For Conversion
AMCIS 2017 PROCEEDINGS(2017)
摘要
This study explores online customer search behavior on a large e-commerce website-Walmart.com. In order to more accurately predict customer purchase conversion based on their search behavior, we adopt a modern machine-learning technique, random forest, as well as logistic regression to develop two computational models. We also integrate information retrieval literature to propose metrics to quantify online consumers' search behavior. Results show that the random forest model performs better with a very high accuracy rate (76%) in predicting customers who will purchase the item they browsed. Among all the predictors, page and session dwell time, user type, click entropy, and click position are the strongest influential factors for the conversion behavior. The findings suggest that, with the enhanced metrics and modeling approaches, search behavior could offer strong cues about customers' purchasing decision. Additionally, the findings also suggest operational implications about how to accommodate and induce the desired search behavior with the e-commerce website.
更多查看译文
关键词
E-Commerce, Search Behavior, Conversion Rate, Computational Models, Random Forest
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络