Pedestrian infrastructure is a key component of walkability and equitable access to public space. However, in Mexico, there is no systematic mapping or national index to evaluate sidewalk quality. This study proposes an automated methodology to construct a Minimum Pedestrian Elements Index (IEMP) based on data from the National Housing Inventory (INV), using unsupervised learning techniques. Binary data on the presence of sidewalks, ramps, pedestrian crossings, trees, street lighting, and pavement type were used for three metropolitan areas in the country: Mexico City, Guadalajara, and Monterrey. Through a combination of data cleaning, robust scaling, dimensionality reduction (PCA), and k-means clustering, we identified pedestrian infrastructure profiles by city. The resulting clusters were then analyzed to assign a score out of 100, reflecting the fulfillment of minimum pedestrian infrastructure conditions. The results show contrasting patterns across cities: Monterrey presents higher homogeneity and more binary outcomes, while Guadalajara and Mexico City exhibit greater urban diversity. This methodology enables the generation of a replicable and comparative diagnosis of pedestrian environmental quality, even in the absence of official records, and opens the door to the design of public policies aimed at improving walkable spaces.