Optimal Data Center Location Selection Using Geographically Weighted Regression and Machine Learning

초록

Data centers serve as critical infrastructure in the digital economy, supporting the growing demand for data processing while contributing significantly to energy consumption and environmental impact. In South Korea, the excessive concentration of data centers in metropolitan areas has exacerbated issues such as power supply imbalances and regional disparities. To address these challenges, this study proposes a methodological framework for optimal data center location selection by integrating machine learning and geographically weighted regression (GWR). Random Forest was employed to identify key factors influencing site suitability, revealing that natural disaster risks (e.g., flood and earthquake risks) and infrastructure conditions (e.g., population density and power supply stability) are critical determinants. GWR was subsequently utilized to estimate region-specific regression coefficients, incorporating local characteristics into the evaluation of location suitability. The analysis identified Cheonan, Gimhae, and Daegu as highly suitable locations, characterized by lower natural disaster risks, accessibility to renewable energy, and favorable infrastructure conditions, thereby ensuring operational stability and sustainability. This study advances the decision-making process by providing a comprehensive framework that considers the interaction between natural disaster risks, infrastructure conditions, and regional characteristics. Furthermore, the proposed methodology has potential applications in other critical infrastructure domains, offering practical insights for achieving regional balance and sustainable data center operations.

키워드