Huff gravity model applications: A data driven machine learning modelfor store close decisions strategy + assessing the efficacy of market share models using this approach on large-scale transactional data + predicting impact on economic results due to diversity in cities

Part one: Numerous studies suggest strategies for identifying optimal sites for new stores and facilities, yet only a few tackle the issue of store closures. Due to the recent COVID-19 pandemic, numerous companies have encountered financial challenges. In this scenario, a frequent approach to avoid losses is to reduce operations by shutting down one or more branch locations. Such choices are typically influenced by the performance of individual stores; consequently, the under-performing locations face the risk of being shut down. This research initially presents a multiplicative variation of the famous Huff gravity model and adds a new attractiveness element to the model. A forward–backward method is employed to train the model and estimate customer reactions and revenue decline following the imagined shutdown of a specific store in a chain. This study examines department stores in nyc by utilising extensive datasets on spatial patterns, mobility, and expenditures. The findings from the case study indicate that the stores suggested for closure according to the proposed model do not necessarily align with individual store performance. This highlights that the performance of a chain stems from the interactions between the stores, rather than merely being an aggregate of their performances viewed as separate and independent entities. The suggested method offers managers and decision-makers fresh perspectives on decisions regarding store closures and is expected to minimise revenue losses associated with such closings. Part two: Extensive research has been conducted on customer patronage behaviour in the marketplace. Engage in the sharing of modelling contexts, a crucial initial stage in the process of effectively modelling and resolving competitive facility location issues. Prior research has utilised surveys to calculate the market share of merchants and determine the elements influencing their appeal for inclusion in diverse hypothetical mathematical frameworks. Emerging developments in the analysis of vast volumes of data facilitating a deeper comprehension of human behaviour and decision-making could potentially yield more realistic assumptions in the development of location models. In this study, we introduce an innovative method to validate the Huff gravity market share model by utilising a comprehensive transactional dataset that captures customer patronage patterns in a specific geographic area. While extensive research has been conducted on the Huff model and its application in competitive facility location and demand allocation, this study represents the initial attempt to verify the accuracy of the Huff model using empirical data. Our methodology facilitates the adaptable implementation of the model across various regions and merchant classifications. The empirical findings demonstrate that the Huff model offers a strong fit for capturing customer purchasing behaviour across various retail settings classifications such as grocery establishments, apparel outlets, fuel stations, and dining establishments. In addition, we perform regression analysis to demonstrate that specific characteristics such as gender diversity and marital status diversity are associated with increased validation of the Huff model. We posit that our analysis, leveraging empirical data, offers compelling support for the validity of gravity-based market share models in the realm of competitive facility location optimisation techniques. Part three: Considerable scholarly attention has been devoted to elucidating the development, advancement, and economic success of urban centres as a whole. However, there is a notable scarcity of empirical data regarding the evolution and economic prosperity of specific neighbourhoods. This study demonstrates that the variety of amenities present in a urban district, derived from publicly accessible points of interest on digital maps, effectively forecasts human movement patterns between urban districts. Furthermore, these movement patterns reliably forecast the economic output of the districts. Furthermore, the variation in consumption patterns or the diversity of transactional activities alongside geographic centrality and population density effectively anticipates local economic development, independent of factors typically considered in analysis, such as population, etc. We construct our models by leveraging geographically tagged purchase data sourced from Istanbul, subsequently verifying the associations through publicly accessible data obtained from Beijing and multiple locations within the United States cities. The findings of our study indicate that the proliferation of goods and services in an urban neighbourhood is the most significant determinant influencing both human mobility patterns and economic development.