Urban parks are essential components of sustainable cities, providing vital health, social, and environmental benefits. Using weekly smartphone-based visitation data for 182 parks in Las Vegas from 2019 to 2022, this study quantifies how the COVID-19 pandemic altered park use and identifies the socio-economic, environmental, and infrastructural determinants of these changes. Park visitation in Las Vegas showed a marked early pandemic decline followed by uneven recovery, with socially vulnerable neighborhoods lagging behind. Ordinary Least Squares (OLS) and Random Forest (RF) models were used to capture both linear and nonlinear relationships. The RF model explained 81% of the variance in standardized visitation (R² = 0.81, RMSE = 0.0415), substantially outperforming the OLS benchmark (R² = 0.24, RMSE = 0.0656). Domain-specific RF models show that socio-economic variables alone achieve an R² of 0.88, compared with about 0.70 for housing, environmental/health, and lighting variables, while demographic variables explain only 0.39, indicating that social vulnerability is the dominant driver of visitation inequalities. Phase-specific analyses further reveal that RF performance increases from R² = 0.84 before the pandemic to R² = 0.87 after it, as park visitation becomes more strongly coupled with socio-economic and health-related burdens. After COVID-19, poverty, uninsured rates, and asthma prevalence emerge as the most influential predictors, while the relative importance of demographic composition and environmental exposure diminishes. These findings demonstrate that pandemic-era inequalities in park visitation are driven primarily by reinforced socio-economic and health vulnerabilities, underscoring the need for targeted, equity-oriented green-infrastructure interventions in disadvantaged neighborhoods.

The rapid expansion of IoT deployments has intensified cybersecurity threats, notably Distributed Denial of Service (DDoS) attacks, characterized by increasingly sophisticated patterns. Leveraging Generative AI through On-Device Large Language Models (ODLLMs) provides a viable solution for real-time threat detection at the network edge, though limited computational resources present challenges for smaller ODLLMs. This paper introduces a novel detection framework that integrates Chain-of-Thought (CoT) reasoning with Retrieval-Augmented Generation (RAG), tailored specifically for IoT edge environments. We systematically evaluate compact ODLLMs, including LLaMA 3.2 (1B, 3B) and Gemma 3 (1B, 4B), using structured prompting and exemplar-driven reasoning strategies. Experimental results demonstrate substantial performance improvements with few-shot prompting, achieving macro-average F1 scores as high as 0.85. Our findings highlight the significant advantages of incorporating exemplar-based reasoning, underscoring that CoT and RAG approaches markedly enhance small ODLLMs' capabilities in accurately classifying complex network attacks under stringent resource constraints.