Network-based quality index aggregation in the retail location problem. A supervised learning approach

In retailing, the location problem is a fundamental strategic aspect. It is usually formalized as a multi-criteria optimization problem to choose the most appropriate spot. A relevant element in the selection is the adequacy of the commercial ecosystem in the vicinity of the location. To account for this criterion, there are different primary indices based on networks that formalize the quality of the available options with regard to the surrounding ecosystem. Previous research suggests that aggregating the different indices using a classifier can improve the quality of these metrics. In this paper, we compare different classifiers to assess their performance in that respect. The analysis has been performed in a context of transfer knowledge and information fusion using data from all the cities in Castile and Leon, Spain. Our results show that the random forest and generalized linear models obtain results significantly superior to other alternatives.