Mağaza Yer Seçimi icin Mekansal Etkileşim Modelleri ile Süpermarket Ciro Tahmini

The goal of this thesis study is to use the customer and store location information for store revenue prediction to support the store location decision. Spatial interaction models are structured and practical analytical formulas and they are often used in store revenue prediction. In this thesis study we focus on creating a new improved spatial interaction model. Basis of our spatial interaction model is insights obtained in the exploratory analysis which we examine factors such as distance, competition, store format through macro and customer level analysis. According to our findings, bigger stores attract more customers from distant areas and distance is a significant variable affecting customer store choice behavior. Based on this finding, we add distance variable to our model and we represent the distance decay value as a function of store size. We observe that this representation of the distance decay value provides improved customer share prediction results. According to our evaluations, a spatial interaction model with store attractiveness values and variable distance decay provides superior individual share of customer spending and store sales prediction results to Huff (1963), MCI (1974) and Competing destinations (1988) models when it is estimated with constrained optimization. We also observe that the spatial interaction model with variable distance decay and store attractiveness values which is estimated with constrained optimization is robust and utilize the information better than conventional spatial interaction models. In this thesis study we also propose a method of store revenue prediction using spatial interaction models and customer loyalty card data when customer panel data and zip code level customer spending are not available.