A Real-Time Monitoring and Analysis Model for Regional Economic Activities Based on Mobile Computing

Abstract

With the rapid advancement of the digital economy, the dynamic and complex nature of regional economic activities has posed significant challenges to traditional low-frequency statistical monitoring methods. Mobile computing, with its capability to capture high-frequency and multi-source data, offers a promising new approach for real-time monitoring. However, current research still faces several limitations: low integration efficiency of heterogeneous data sources, insufficient fusion of spatiotemporal features, and significant interference from noise. Traditional statistical models, reliant on low-frequency sampled data, often suffer from lag and sampling bias. Although machine learning methods have improved predictive accuracy, models such as long short-term memory (LSTM) lack the ability to capture spatial heterogeneity; conventional denoising algorithms struggle to handle complex noise patterns; and many studies fail to fully explore the spatiotemporal coupling of economic activity. To address these issues, this study proposes a real-time monitoring and analysis model for regional economic activities based on mobile computing. The model consists of five core modules: (1) a data acquisition and preprocessing module for real-time integration and outlier detection across multiple data sources; (2) a denoising module based on the rime ice optimization (RIME) algorithm, which enhances robustness against noise through soft frost search and hard frost penetration mechanisms; (3) a spatial feature extraction module using graph attention networks (GAT) to model inter-regional economic relationships and capture spatial spillover effects; (4) a temporal feature extraction module based on LSTM to uncover long-term temporal dependencies; and (5) a prediction output module that fuses spatial and temporal features for accurate forecasting of economic activities. The innovation of this model lies in its optimized denoising process through the RIME algorithm and the deep integration of spatiotemporal features via GAT-LSTM. It supports real-time data input and dynamic prediction, providing an intelligent tool that transforms regional economic governance from “post-event analysis” to “real-time perception.” This contributes to more precise policymaking and more efficient resource allocation.