# Explaining urban street perception inequities between residents and tourists using interpretable machine learning

## TL;DR

This study uses machine learning to understand how residents and tourists perceive urban streets differently, aiming to improve inclusive urban design.

## Contribution

The novel contribution is an interpretable machine learning framework that identifies perceptual differences between residents and tourists using visual and environmental features.

## Key findings

- Tourists prioritize symbolic and aesthetic cues, while residents focus on functional and comfort-related features.
- Visual elements like vegetation and spatial openness affect perception differently for residents and tourists.
- The framework reveals actionable insights for equitable urban street design.

## Abstract

Understanding how different social groups perceive urban streets is essential for inclusive and sustainable urban design. This study proposes an interpretable and scalable machine learning framework that integrates Street View Images with subjective evaluations to examine perceptual differences between residents and tourists. Using data from Xi’an’s historic Mingcheng District, we collected perception ratings across five dimensions-safety, comfort, convenience, pleasure, and sociability-and analyzed how visual and environmental features shape these perceptions. The framework combines predictive modeling and explainable analysis to uncover both linear and nonlinear drivers of perception. The results show that tourists are more responsive to symbolic and aesthetic cues, while residents emphasize functional and comfort-related features. Key visual elements such as vegetation, building facades, and spatial openness exert different effects on the two groups. By revealing these perceptual disparities, the study provides actionable insights for perception-informed and equitable street design strategies that better address the needs of diverse urban users.