A paper titled “Comparing Aggregation Methods for Road Safety Analysis Using Representation and Machine Learning Techniques” authored by Simone Paradiso, Apostolos Ziakopoulos, Haris Sideris and George Yannis has been published in Data Science for Transportation. This Paper proposes a refined way to aggregate smartphone-based telematics data into nodes and edges, using a dataset of multiple trips represented by geospatial coordinates. The refined node method limits data points to edges directly connected to the buffer’s origin node, reducing node-level variability by roughly 7-50% across features. Furthermore, principal component analysis and an autoencoder were used to study the reconstruction of latent representations, and Global Moran’s I values were lower, suggesting reduced artificial similarity and more distinct local variation. Moreover, a random forest classifier linked telematics features to crash occurrences, and the authors conclude that the proposed approach gives a cleaner, less redundant representation, yielding a clearer crash risk signal. This proposed approach suggests a cleaner and less redundant representation of the underlying relationships, yielding a clearer crash risk signal.