Abstract: Some major cities have formulated relevant policies and planning strategies targeting at urban heat island effect. However, such policies and strategies seldom consider public space visiting behavior, citizen walking behavior and three-dimensional (3D) characteristic of high-density cities. Taking the small public spaces in the Central and Western District and Wan Chai District in Hong Kong as an extreme example of volumetric urban design, this research uses 3D spatial design network analysis (3D sDNA) to measure the accessibility and traffic potential of public spaces based on a cognitive pathfinding mechanism. Additionally, by virtue of unsupervised machine learning, the research investigates the relationships between such factors as size, location visibility and design quality of the public spaces in Hong Kong, finding that the design quality of such public spaces improves with the increase of their area and visibility. In combination with small public spaces, urban ventilation and daily routes most frequently used by pedestrians, the research outlines a design sketch for a proposed “cool network” in Wan Chai, with a view to enabling the “cool network” to serve all age groups with different occupations therein. Through both quantitative and qualitative discussions, the research explores the complexity of accessibility for pedestrians based on which a logical space design intervention method was proposed, to increase both the usage frequency and the coverage of user groups of public spaces while providing a powerful support for planning and design practice regarding public spaces in high-density cities.