Visualizing and Analyzing Human Activity Patterns through Graph-Based Methods

Saima Ahmed Rahin, Bo Hui, Wanwan Li

Abstract: In recent years, the proliferation of wearable sensors and mobile devices has enabled the collection of large-scale spatiotemporal data, providing unprecedented opportunities to analyze human activity patterns. This study presents a novel approach to visualize and analyze human activity patterns using graph-based methods. Utilizing a dataset comprising timestamped geographical coordinates and labeled activity data, we construct a graph where nodes represent individual data points, and edges denote temporal proximity. By employing Graph Neural Networks (GNNs), we effectively capture the intricate spatiotemporal relationships inherent in the data. Later, we compared three other models GCN, GAN, and GraphSAGE, and found that GCN performs better for cluster separation and GAT shows best in terms of training loss.