A novel O2O service recommendation method based on dynamic preference similarity

Recent technological advancements have enabled an increasing number of consumers to select services from online platforms and utilize them in offline stores, a model known as online-to-offline (O2O) e-commerce. This emerging model has garnered significant attention from both business and academic communities. However, with the rapid growth of O2O services, consumers face challenges in selecting services that align with their preferences from a vast array of options. To address this issue, this paper proposes a novel O2O service recommendation method based on dynamic similarity estimation (ReDPS). The dynamic similarity is calculated by tracking changes in consumer preferences over time, providing a more accurate and robust measure of consumer relationships. We validate the ReDPS method using both the Dianping dataset and the publicly available Yelp dataset. Experimental results show that: 1) ReDPS significantly outperforms classical and state-of-the-art recommendation methods, with its effectiveness improving over longer time spans of consumer feature data. 2) Consumer preferences are more strongly influenced by variations in service categories and geographical locations over time than by changes in service evaluations, though all factors are important, and consumers of the same gender tend to exhibit similar preferences. 3) Optimal parameter configurations for ReDPS are identified through the experiments.