Improving recommender systems with an intention-based algorithm switching strategy

Modern e-commerce websites are equipped with hybrid recommendation systems aiming at bringing novelty and diversity to consumers. However, mobilizing several recommendation algorithms simultaneously not only incurs unnecessary computation costs, but also jeopardizes consumers' shopping experience due to excessive information load. Hence, recommending less but better (more relevant) items is critical, especially when consumers depend more and more on mobile devices, whose screen is much smaller. In this paper, we present a switching hybrid strategy capable of selecting recommendation algorithms according to consumers' instantaneous intention. Compared with a benchmark system which simultaneously uses all algorithms, our system achieved higher performance in terms of item view and consumption while sending less items, though both systems are empowered by the same recommendation algorithms. Meanwhile, the interface of our system is more concise and user-friendly. The result indicates that the intention as an important context factor can be used to enhance the performance and consumer experience of e-commerce recommender systems.