Desarrollo de un modelo de recomendación de rutas personalizadas para grupos de ciclistas empleando sequence-aware recommender systems.

Abstract

In recent times, there has been a notable increase in the adoption of technologies and interactive platforms within the realm of physical fitness and sports. This trend is exemplified by the emergence of products such as Strava, Garmin Connect, and ®Fitocracy. The competitive dynamics within this sector have driven progress in functional features, resulting in the development of more user-friendly applications that reshape the sociotechnical landscape of sports practices. At the same time, despite the widespread prevalence of recommendation systems in sectors like e-commerce and entertainment, their integration into new domains, particularly the sports industry, remains relatively limited. This study presents the creation of a Group Recommender System (GRS) designed for per-sonalized cycling routes, employing Sequence-Aware Recommender Systems. The main objective is to improve the user experience by providing route suggestions to groups of cyclists, encouraging the discovery of new routes. Anchored in the Design Science Research (DSR) framework, the project lays the foundation for an extensive conceptual framework governing the design and construction of artifacts. Experimental findings highlight a positive correlation between normalized discounted cumulative gain quality (NDCG) and user group size, underscoring enhanced performance in larger groups due to increased diversity. Additionally, in model training with the GoSS-Rec algorithm, the Prod2vec model consistently outperforms alternatives in diversity and novelty metrics for both small and large groups. This emphasizes the potential of the GoSS-Rec algorithm with Prod2vec to excel in providing diverse and novel recommendations across groups of varying sizes, contributing to the evolution of recommendation systems in nontraditional domains.