The exponential growth of Internet of Things (IoT) technologies has intensified the demand for efficient computing solutions to handle the massive amount of data generated by connected devices. Edge computing, as a paradigm, offers a promising solution by decentralizing computations closer to data sources. This study introduces a novel framework that leverages support vector machines (SVMs) for dynamic resource allocation and load balancing in edge computing environments. Experimental evaluations demonstrate that the SVM-based framework achieves significant performance improvements over heuristic-based, clustering-based, and other machine learning approaches. The results reveal that the SVM framework reduces the total latency by 14.2% and 21.6% compared with heuristic and clustering methods, respectively, and outperforms models such as K-nearest neighbors, random forest, and neural networks by achieving the lowest latency (1.803125), best load distribution (0.073357), and highest cost efficiency (0.877428). These findings highlight the SVM model's ability to optimize resource utilization, reduce task completion times, and improve system adaptability. Its low computational overhead and predictive capabilities make it particularly suitable for latency-sensitive applications, such as healthcare IoT and autonomous vehicles. Furthermore, the study discusses limitations and proposes hybrid model integrations to address scalability and real-time adaptability for future research.