Application of self-organising maps to analysed bike-sharing usage in large cities: A case study of Seville

This study provides a quantitative analysis of the relationship between Seville’s public bicycle-sharing system usage and the city’s environmental conditions, employing an unsupervised artificial intelligence method. The dataset spans from 2018 to 2023 and combines operational service data (daily rentals, turnover) with weather data (temperature, precipitation, humidity), including the heat index as a measure of thermal comfort. The method is based on self-organising maps (SOM), implemented through a neural network with a 12 × 12 hexagonal grid. After 100 iterations of unsupervised training, hierarchical clustering was applied to the neuron weight vectors, resulting in 8 clusters representing distinct patterns of usage and environmental conditions. The results indicate that segmentation is strongly influenced by weather variables, showing a negative relationship between service demand and precipitation and a nonlinear relationship with thermal sensation, which discourages system use at extreme levels. The proposed model provides a practical tool for operational planning and demand forecasting, enabling system management optimisation based on weather forecasts. Additionally, its application could benefit other cities with similar mobility systems.