Analyzing the Evolution of Edge Computing Technologies Through RI-Optimized LDA Topic Modeling and Ensemble Forecasting

Abstract

The rapid growth of Internet of Things (IoT) technologies has elevated edge computing among distributed computing models, necessitating a thorough innovation analysis. In this study, we conducted a systematic analysis of 1,616 edge computing patents (2015-2024) in the US-predominant patent environment using an improved topic modeling approach that fills important gaps in patent landscape research methodology. We used the Relative Independence (RI) model for optimal topic selection, balancing topic internal consistency and separation. Our analysis showed that K=4 is the optimal number of topics (RI score=0.4552), defining four technological areas: Vehicular Edge Intelligence (24.75%), Edge-Cloud Orchestration (31.50%), Edge-based Sensing (18.69%), and Multi-Access Edge Network (25.06%). Age-Normalized Citation Impact analysis identified technologically important patents by evaluating citation speed rather than magnitude. Topic-topic network analysis showed Edge-Cloud Orchestration as the key knowledge hub with the highest PageRank centrality (0.2878) and in-degree (12,538). For forecasting, we proposed a hybrid ensemble model combining Bass Diffusion, Gompertz, and Logistic growth models with LightGBM machine learning. Gompertz models outperformed in fit for three of four domains (MSE from 6.51 to 14.21), while Bass Diffusion best modeled Topic 1’s adoption process (MSE=7.72). Forecasts show Edge-based Sensing growing fastest with a CAGR of +23.5%, followed by Edge-Cloud Orchestration at +19.1%, while Vehicular Edge Intelligence and Multi-Access Edge Network show moderate growth at +9.2% and +9.1%, respectively. U.S. dominance at 95.1% indicates a concentrated innovation focus and the need for global technology transfer.