Spatiotemporal Evolution Analysis of Social Media Tourism Hotspots Based on Pulse-Coupled Neural Network and Memristor Cross Array

Abstract

In the spatiotemporal evolution of social media tourism hotspots, feature aliasing leads to computational bottlenecks and low analysis efficiency. Therefore, this study investigates spatiotemporal evolution analysis of social media tourism hotspots based on pulse-coupled neural networks and memristor cross arrays. A hierarchical memristor cross array computational architecture is constructed to adapt to four-dimensional spatiotemporal tourism data and achieve in-situ parallel computation. The dynamic model of the pulse-coupled neural network is reconstructed, and nested coupling operations are introduced to complete pulse coding and unsupervised extraction of multimodal tourism features from social media. Combined with spatial correlation analysis and temporal inference algorithms, tourism hotspot identification, heat quantification, and spatiotemporal evolution law analysis are completed. Experiments show that the research method achieves a spatiotemporal evolution path matching accuracy of 0.975 under a 24-hour daily observation window, and achieves a high-stability classification effect with single-class recognition accuracy exceeding 95% for four types of cultural tourism scenarios. Under a 2048-dimensional high-dimensional feature input scenario, the inference latency is compressed to 18.7ms, and the computation speedup ratio reaches up to 8.7 times, improving the efficiency of spatiotemporal evolution analysis of social media tourism hotspots.