Modeling the Time Spent at Points of Interest Based on Google Popular Times

Location-based applications are increasingly popular as smartphones with navigation capabilities are becoming more prevalent. Analyzing the time spent by visitors at Points of Interests (POIs) is crucial in various fields, such as urban planning, tourism, marketing, and transportation, as it provides insights into human behavior and decision-making. However, collecting a large sample of behavioral data by using traditional survey methods is expensive and complicated. To address this challenge, this study explores the use of crowdsourcing tools, specifically Google Popular Times (GPT), as an alternative source of information to predict the time spent at POIs. The research applies a robust regression model to analyze the data obtained from GPT. The popularity trends of the different POI categories are used to indicate the peak hours of the time spent in the city of Budapest. Non-spatial parameters such as the rating, the number of reviewers, and the category of the POIs are utilized. Furthermore, a Geographic Information System (GIS) is applied to extract the spatial parameters such as the security and safety levels, the availability of car parking, and public transport (PT) stations. The robust linear models are statistically significant based on the p-values, thus indicating a strong relationship between the independent variables and the time spent at POIs. The weekday and weekend models present 69.5% and 73.9% of the variance in the time spent at POIs, respectively. Furthermore, it is demonstrated that the visitors’ behavior is strongly affected by the category of the POIs variable. This study shows how GPT can be utilized to better understand, analyze, and forecast people’s behavior. The solution presented in this study can serve as an essential support of activity-based models, where the time spent is a crucial parameter for scheduling and optimizing activity chains.