A Neural Network Model for Color Element Data Analysis for Urban Spatial Environment.

Area of Science:

• Computer Science
• Urban Planning
• Data Analysis

Background:

• Urban spatial environments generate complex color element data.
• Analyzing this data is crucial for understanding urban characteristics and planning.
• Existing methods may not fully capture the intricate relationships within spatial data.

Purpose of the Study:

• To develop a Convolutional Neural Network (CNN) model for effective color element data analysis in urban spatial environments.
• To investigate and integrate high-order structures using a motif-based graph autoencoder.
• To enhance graph representation learning with redefined similarities and awareness mechanisms.

Main Methods:

• Construction of a CNN model for color element data analysis.
• Proposal of a motif-based graph autoencoder integrating first- and second-order similarities.
• Implementation of an efficient graph transformation and a primary awareness mechanism.
• Classification validation using a support vector machine with CNN features.

Main Results:

• The proposed Cen GCN_D and Cen GCN_E variants show superior performance in node classification, link prediction, clustering, and network visualization.
• Classification accuracy using CNN features ranges from 91.4% to 95.2% with 450 training images.
• Accuracy consistently exceeds 90% when the training set surpasses 300 images, indicating stable trends.

Conclusions:

• The developed CNN model provides a robust framework for analyzing urban spatial environment color data.
• The motif-based graph autoencoder effectively captures complex data structures.
• The study achieves fast classification prediction, offering a valuable tool for urban environmental studies.