Enhancing Reverse Geocoding With Weather Data: Modeling Human Check‐In Behavior in California and New York for Smart Cities

ABSTRACT This paper presents a novel reverse geocoding approach that integrates multidimensional weather data into spatial reasoning, providing a context‐aware and behaviorally driven alternative to existing solutions. By emphasizing the point of intersection between environmental factors and human mobility, it proposes a framework for enhancing reverse geocoding by modeling check‐in behavior influenced by weather patterns. The key innovation lies in incorporating environmental conditions—specifically air temperature, humidity, wind speed, and cloudiness—into the geocoding process to model how weather influences human check‐in behavior. While conventional methods rely primarily on spatial proximity, they often ignore such behavioral contexts. In our approach, 1 year of Swarm check‐in data from California and New York were used to create weather semantic signatures for 599 venue categories, representing check‐in likelihood under different weather conditions. These signatures were integrated into several mathematical distortion models to adjust spatial relationships, with the linear model showing the highest accuracy. To further enhance the model, Particle Swarm Optimization (PSO) was applied to determine optimal parameter weights. The proposed method was evaluated on unseen temporal data to assess the model's generalizability. The Mean Reciprocal Rank (MRR) improved by 14.90% in California and 13.40% in New York, while PSO optimization raised these to 18.48% and 15.37%, respectively. The First Position index also improved significantly by 76.43% and 83.04%. These findings highlight the effectiveness of the proposed method for enhancing reverse geocoding for urban applications.