A machine-learning-based approach to online transaction analysis for personalized customer services

With the amount of online transactions surging, online marketplaces are generating millions of transaction data and customer profiles. Providing personalized customer service is becoming a rewarding yet challenging task for online sellers and stores as online customers grow both in number and in diversity [1]. In this paper, we propose a machine learning based pipeline to analyze and visualize online customer transaction data. We perform data cleaning and feature extraction on a collection of online transaction data. We then implemented dimension reduction and clustering algorithms on the feature data to group customers into segments. Finally, association rules are applied to further analyze transaction data in each customer segment.