Optimal Risk-aware POI Recommendations during Epidemics

Nina Yanin, Manos Papagelis

Abstract: The movement of people can influence the spread of diseases, especially in populated areas. While measures like quarantine can curb disease spread by restricting the movement of those infected, they come with socioeconomic consequences. Furthermore, not everyone might adhere to these restrictions, undermining their effectiveness. A more effective strategy is to educate people on the risks tied to their movement habits and recommend safer options. In this research, we introduce the problem of optimal risk-aware point-of-interest (POI) recommendations during epidemics, where people get recommendations on what POI to visit that reduces the risk of getting infected. The risk of infection at a POI is modeled based on its capacity and visit patterns over time. Then, we present a method that provides personalized recommendations which, when universally adopted, the overall risk is minimized. Unlike existing strategies, our method considers simultaneous user requests made in the same time period, which might influence the relative risk at POIs. An extensive evaluation was conducted, using real-world data coming from three major cities in Canada, which showed that our method outperforms the current state of practice method and other sensible baselines, on varying settings. Specifically, our method presented a decrease in the relative added risk of infection by 99.87%, 71.56% and 61.54% at each city, respectively. We also examined how the optimal solution is impacted if only a specific portion of the population follows the recommendation. Our optimal risk-aware recommendation method has the potential to reduce infection risk by promoting responsible behaviors within communities.