In this research, we develop a “sensory score” that quantifies the sensory characteristics of cities using SNS data, and then built a prediction model for urban images by combining urban data and SNS data. As a result, although the impact of using SNS data is limited to a slight change in terms of the improvement of prediction accuracy, the prediction model using SNS data can express the localized distribution. Thus, it can be pointed out that using SNS data is more effective than the model that using only objective urban data in terms of evaluating local trends in urban images.
