Validating Crowdsourced Building Data: A Statistical and Expert Approach

The SMILE project explores the use of machine learning to generate updated building exposure layers by integrating remote sensing imagery, census data, and validated crowdsourced information. Crowdsourced data are collected through targeted initiatives involving trained students and citizens. To facilitate these activities, a web-based multimedia platform (https://smile.mi.imati.cnr.it) was developed to guide users through data collection, manage workflows, and store georeferenced information and images in a structured database, which currently includes survey forms on 4,100 buildings, mostly located in Northestern Italy.

A key goal of the study is to validate the collected data and assess their potential use to enhance existing building exposure datasets. Approximately the forms for about 400 buildings located in Udine, filled in by high school students via the platform, were reviewed by experts. Comparing expert-validated and student-collected data enabled identification of potential issues in survey design and allowed for a statistical assessment of data quality and reliability. The validation approach integrates rigorous statistical techniques, including summary statistics, cross-correlation analyses, and dissimilarity measures, with visualization methods to support the interpretation and communication of complex datasets.

We acknowledge the PRIN 2022 PNRR project SMILE “Statistical Machine Learning for Exposure development” (code P202247PK9, CUP B53D23029430001) within the European Union-NextGenerationEU program.