Open Access Thesis
Store location--Decision making; Machine learning;
Businesses use GIS software to build spatial business models to have possibility to analyze customers, competitors or markets not only in terms of finance, but also in terms of their behavior in space. Specifically, geomarketing methods are used to maximize profit when searching for places to open new factories, shops, restaurants, or expand chain of cafes. The main geomarketing approach is to identify the optimal location based on socio-economic data, as well as on the criteria that are necessary for this type of business (suitability model).
Traditionally the Maxent model, a type of an ecological niche model (ENM), is used to forecast the distribution of different species of animals or plants in biological science. This machine-learning algorithm uses data from points where the phenomenon in question has already been found and selects locations with similar characteristics. Thus, the location is not determined by the expert choice, based on the conditions and factors that are perceived as necessary or beneficial, but is identified on the basis of the machine learning algorithms driven by statistical assessment of existing locational information for a given species.. However, his modeling approach has applications beyond species distribution and has already been used to examine other phenomena,such as optimal locations of wind turbines.
This study tests the opportunities of using Maxent for geomarketing applications, specifically determining the ‘best’ locations for coffee chain restaurants using Biggby locations in Michigan State as an example.The results indicate that Maxet and EMN principles could be potentially useful for geomarketing applications. The resultant model used a very limited number of variables but was able to demonstrate in principle the utility of the method. Also, 2 cell scales were used to compare how different the results of the models can be when using different scales. The main variables that were used in the study are based on those indicators that the Biggby coffee shop network uses directly to find new locations to expand its chain.The main limitations of this analysis are the lack of contact and support from Biggby, as well as the inaccessibility of certain types of data, as their traffic intensity. Another limitation in the work was the difficulty in processing a large amount of data, which was formed due to the fact that initially the analysis was to be applied to 4 states.
Year of Submission
Master of Arts
Department of Geography
Andrey N. Petrov, Chair, Thesis Committee
1 PDF file (xi, 163 pages)
©2020 Petr Grin
Grin, Petr, "Utilization of machine learning algorithms to support retail chain store location decisions" (2020). Dissertations and Theses @ UNI. 1048.