Commercial Space Classification and Prediction with Machine Learning and Multiple Data Sources: A Case Study of Milan

As cities evolve, urban commercial spaces become increasingly complex and diverse, requiring effective development and management to optimize land use and economic efficiency. This paper introduces a machine learning-based method for recognizing and predicting commercial spaces, addressing limitations of traditional approaches, such as reliance on small sample sizes and limited scope of urban factors, subjectivity in categorization, and static interpretations that fail to capture dynamic urban changes. Using data from 26 commercial activity types and 17 urban factors, Milan serves as a case study where commercial spaces are clustered into six typical categories. A random forest model is then trained to predict these categories, achieving a top-3 accuracy of 88.7%. The methodology demonstrates its practical application by predicting the impact of Milan’s Metro Line 4 on future commercial development, identifying opportunities and risks for various spaces. This approach provides valuable insights for Milan and is adaptable to other urban contexts.