Intent-based recommendation for B2C e-commerce platforms

Recommendation systems are a well-established component of business-to-customer (B2C) e-commerce websites. The goal of a recommendation system is to increase sales by accurately predicting additional items that a customer will buy, but has not yet considered, during his or her online shopping session. Some companies, such as Amazon.com, consider the efficacy of their recommendation systems a critical competitive advantage. As competition intensifies, many online retailers seek more effective methods to anticipate customers' needs and desires during their online experience. In this paper, we present a novel approach for increasing the conversion rate from browsing to buying customers. Our method makes a behavior-based inference of a customer's propensity to purchase from a product category, and then executes a dynamic recommendation plan that is conditioned on the propensity levels. The approach uses a hidden Markov model, classic recommendation algorithms, and business rules.