Researchers Propose A Graph-Based Machine Learning Method To Quantify The Spatial Homogeneity Of Subnetworks

This Article Is Based On The Research Paper 'Quantifying spatial homogeneity of urban road networks via graph neural networks'. All Credit For This Research Goes To The Researchers 👏👏👏

Please Don't Forget To Join Our ML Subreddit

Purdue University and Peking University researchers recently completed a study employing machine-learning methods better to comprehend road networks in cities throughout the world. Their research, which was published in Nature Machine Intelligence, details the findings of a data-driven investigation of road map-related data collected in 30 cities around the world.

According to the researchers, the urban road networks (URNs) serve as a city’s economic engine. They are formed by many socioeconomic elements (including population, economics) and urban development history. They are essential for human mobility, healthy movement, the propagation of biological viruses, and pollution creation. Traditional road network measurements based on simple parameters, on the other hand, only provide a rough description of URNs.

Previous research has suggested that the spatial homogeneity of URNs follows a pattern. Graph neural networks (GNNs) is an advanced graph-based machine learning methods commonly employed in computer vision and natural language processing applications that could capture these patterns. GNNs can learn network representations from very vast volumes of data. The research team employed GNNs to evaluate over 11,790 URNs in 30 cities worldwide to predict a new parameter called network homogeneity in their research.

The researchers explained that many urban planners had examined cities using case-by-case methodologies. They had sought to use the power of machine learning technology and knowledge from huge data to gain a worldwide understanding of the urban system, which included cities in both developed and developing countries. However, utilizing worldwide data, quantitative comparisons of the urban environment among cities are limited.

The researchers separated all of the URNs in their dataset into two portions before completing their analyses: the “hidden region” and the “observed region.” Following that, they trained the GNNs to learn the structure patterns of road networks in the observed regions, allowing them to predict the network structure in the hidden region.

Source: https://arxiv.org/pdf/2101.00307.pdf

The researchers defined the metric they studied, “network homogeneity,” as the F1 score of the model’s performance in predicting the concealed region in the URN data in their article. F1 scores represent classifiers’ precision and recall accuracy or machine learning systems. A higher F1 score meant that the model was more likely to successfully infer the hidden region from the seen region in the context of the team’s research. It also meant that URNs were more homogeneous.

Top-down urban planning policies often govern the evolution of road networks in large cities. These policies could be evaluated and compared using the researcher’s graph neural network-based model and metric.

The researchers’ analysis revealed links between URNs, a country’s gross domestic product, and population growth and proposed a statistic that could aid in the evaluation of urban planning initiatives. The findings support the intimate relationship between human activities and urban environments.

The findings of the research team helped them comprehend the complex interplay between numerous components in the urban system. Their study is unusual in that it combines machine learning with urban science: the F1 score is a typical machine learning metric, while homogeneity characterizes the network structure of URNs.


According to the research group, they were the first ones to look at whether machine learning algorithms might meaningfully study URN systems. Their findings show that advanced machine learning algorithms might be utilized to collect rich data on socioeconomic aspects and city evolution over time.

Imagine walking around the corner of a street and forecasting what the following block will look like just by looking at the neighborhood you just passed. When traveling in a new city, you may feel highly acquainted with the surrounding area and believe it is identical to another place you are familiar with. This could occur in various cities within the same country or across multiple countries. Interestingly, this occurrence isn’t random but rather can be explained by simple elements like intracity and intercity homogeneity and can be traced back to the urban planning culture of Europe, North America, and Asia, from ancient to modern cities.

The neural graph networks constructed by this group of academics could eventually be utilized in various countries throughout the world to compare cities, evaluate policies, and summarise activities. Surprisingly, the model presented in the latest research might be extended to investigate more significant urban regions and examine changes over more extended time periods.

The researchers shared that other studies can look into road networks in cities across the globe and of various sizes, whereas their analysis looks at 30 large cities. Furthermore, they can assess and evaluate the homogeneity of a road network across time. The homogeneity theories of street views, land use, and other infrastructure networks could be constructed in addition to road networks. There is a lot more chance to learn more about complicated urban processes.

The team’s new study is one of the first to use advanced graph neural networks to investigate URNs. The researchers hope to improve their model and apply it to more data in future studies to learn more about URN homogeneity. They are now conducting research into URNs and their links to various socioeconomic characteristics.

The researchers want to employ machine learning models to look at other forms of data collected in urban contexts and look at the evolution of URNs through time. For example, they want to examine street view photographs, mobility interactions, and internet browsing data at the same time to find more complicated patterns that affect people’s lives in cities. For example, this could aid in a better understanding of social inequality and regional poverty.

Finally, the researchers want to undertake more research into AI’s overall potential in urban science. Their study may motivate other research organizations to use machine learning in the field of urban science, leading to new insights into the history and evolution of cities around the world.

Source: https://techxplore.com/news/2022-05-graph-neural-networks-spatial-homogeneity.html

Paper: https://arxiv.org/pdf/2101.00307.pdf