Utilizing Skip-Gram for Restaurant Vector Creation and Its Application in the Selection of Ideal Restaurant Locations

Abstract

Restaurant Site Selection (RSS) plays a pivotal role in the success of launching a new restaurant. The core elements of RSS encompass foot traffic and the consumption capacity potential at prospective sites. Previous studies often relied on data gleaned from social media or the Internet, utilizing statistical or machine learning methods to predict foot traffic. Nevertheless, amassing comprehensive data on foot traffic and consumption capacity proves arduous. Multiple factors, such as MRT flow, bus traffic, and business districts, contribute to foot traffic, rendering data collection complex. Similarly, quantifying consumption capacity involves variables like salary and the habits of residents and workers in the vicinity, posing data collection challenges. In contrast to prior work, this study derives proximity insights from numerous restaurant types and their locations. Employing the n-skip gram mechanism from natural language processing, restaurant vectors are generated for each restaurant type. These vectors subtly encapsulate information about foot traffic and consumption capacity. Subsequently, the algorithm utilizes these Restaurant Vectors to recommend optimal restaurant locations. Performance assessments confirm that the generated Restaurant Vectors effectively encompass features related to foot traffic and consumption capacity.