Article Preview
TopDecision-making on facility location-allocation is usually conducted based on optimization methods. The existing pieces of literature include single-objective and multi-objective optimization models. For single-objective, the authors proposed models to improve either monetary or non-monetary criteria such as minimized transportation cost and solved by exact algorithm (Horner et al., 2018), minimized total cost and solved by genetic algorithm (Praneetpholkrang & Huynh, 2020), minimized number of shelters and solved by exact algorithm (Ozbay et al., 2019). In aspects of multi-objective, dealing with several and opposed objectives is difficult but it is a necessity when conducting decision-making in practice (Nikulin et al., 2018). Some of prior works minimized demand weighted distance, number of shelters and employed epsilon constraint method for a trade-off (Görmez et al., 2011), maximized demand coverage and minimized demand weighted distance, then solved by epsilon constraint method (Chanta & Sangsawang, 2012), maximized demand coverage while minimized operation cost and used weighted goal programming to solve the proposed model (Hallak et al., 2019). In addition to shelters, the researchers also formulated the models to determine proper locations for setting distribution centers (Burkart et al., 2017), emergency facilities (Barzinpour & Esmaeili, 2014; Mejia-Argueta et al., 2018), and healthcare centers (Miç & Koyuncu, 2019).
Other than optimization, the researchers incorporated ML with other applications to determine facility location-allocation. Neural network, support vector regression, linear regression, and boosted regression were combined with automated web GIS to predict the hotels’ locations (Yang et al., 2015). Moreover, ANN was also integrated with fuzzy AHP to select locations for establishing convenience stores (Kuo et al., 2002). The existing pieces of study implied the opportunity of ML in solving location-allocation problems, and still room for further studies.