Exploring Sentiment Analysis and Segmentation of Online Marketplace Data

The purpose of the project was to provide an online marketplace platform with a foundational model for building a personalized stakeholder (seller) experience that would enable sellers to better understand their online presence. This case study first accomplished the creation of an entity-relationship diagram to highlight business/technical data relationships and to increase understanding needed to accomplish the subsequent case study objectives. Using this foundational knowledge, sentiment analysis of customer order reviews was conducted to offer insights on operational enhancement and refinement to an online marketplace seller’s storefront. Various models, specifically the BERT Model, Logistic Regression, and Naive Bayes, were compared with relevant data processing techniques to select the best method. Lastly, customer segmentation was conducted using the K-Means Clustering algorithm extended by Recency-Frequency-Monetary (RFM) analysis to understand customer shopping behavior. Following this, targeted email campaign suggestions per customer segment grouping were provided for the purpose of enhancing a seller’s marketing presence and increasing customer engagement and lifetime value.