Identifying Arbitrage Opportunities in Retail Markets using Predictive Analytics

Abstract

While the body of literature on arbitrage opportunity identification is well established in the financial, energy, and real-estate markets, research on arbitrage in the retail marketplace is relatively limited. As arbitrage opportunities in this context differ inherently from other markets, we propose a data analytics framework to identify such an arbitrage by leveraging a machine learning model to predict the optimal purchasing point from the price movement. Our predictive approach is enhanced by incorporating user-generated content which demonstrates its informative power. Overall, the enhanced model attains the precision rate of more than 90 percent while the recall rate is higher than 80 percent in a cross-validation test using the data collected from Amazon Marketplace. In addition, we conduct a field experiment to verify the external validity of the model in a real-life setting. The result shows that our model is capable of generating as much as a 113.31% profit margin.

Keywords: arbitrage, user-generated content, predictive analytics, field experiment, random forest