Business point of interest recommendation with reinforcement learning

The importance of location for business success cannot be overstated. Existing approachesto the business location selection problem often involve creating extensivelytuned models specific to the geographical and economic climate being analyzed, andthus suffer from limited generalization across diverse scenarios. This thesis proposesa novel Deep Q-Learning framework for business location recommendation that canbe trained in one geographic area and applied to another without requiring furthertraining or tuning. Comprehensive experiments on real-world data demonstrate thesuperior generalizability of the proposed recommendation framework, outperformingthe well-established Huff gravity model by 15.33% in profits, with an average profitrealization of 78.02% compared to the best-case scenario. Empirical results indicatethat variation in training data must be as high as the variation in the test datafor the framework to be successfully applied to other locations despite discrepanciesbetween the characteristics of the cities. The proposed approach offers a highly generalizableand easily applicable solution to the business location selection problem,providing a strong alternative to gravity-based models.