Enhancing Urban Data Analysis: Leveraging Graph-Based Convolutional Neural Networks for a Visual Semantic Decision...

Area of Science:

• Urban Planning
• Data Science
• Computer Vision

Background:

• Modern urban environments generate vast amounts of data from diverse sensors, creating challenges in data integration, visualization, and utilization.
• Effective urban planning requires optimal site selection for commercial activities, public utilities, and infrastructure reuse, necessitating advanced data analysis techniques.

Purpose of the Study:

• To evaluate the effectiveness of Convolutional Neural Networks (CNNs) with graph-based urban data representation for urban planning decision support.
• To compare the performance of CNNs against previous methods, specifically random forests, using a consistent dataset.

Main Methods:

• Utilized a graph-based representation for urban data, leveraging its inherent visual characteristics.
• Employed Convolutional Neural Networks (CNNs) for classification tasks, inspired by their image-based data processing capabilities.
• Compared CNN performance against random forests on a standardized urban dataset for location selection.

Main Results:

• The CNN approach demonstrated improved performance across several key indices compared to the random forest baseline.
• The combination of CNNs and graph-based urban data proved effective for decision support in urban planning.
• Results indicate a promising potential for this methodology in addressing complex urban planning challenges.

Conclusions:

• CNNs integrated with graph-based urban data offer a superior approach for urban planning decision support.
• This method enhances the ability to select suitable locations for urban development and infrastructure.
• The findings suggest a significant advancement in utilizing complex urban data for informed planning.