AI-Based Load Balancing Using Decision Tree Regressor for Parallel Matrix Computation in Cloud Environments

Cloud computing is an evolving technology of current information systems that supports dynamic sharing and elastic provision of resources and services. With increasing demands for computational resources, efficient workload assignment has become an important challenge. Current load balancing methods based on traditional approaches fail to suit dynamic server performance and contribute to the inefficient utilization of available resources, latency, and delays. In response to this challenge, this paper suggests an AI-driven load balancer based on a decision tree regressor to dynamically control task allocation within a parallel cloud system. The system operates to handle computationally heavy tasks, i.e., matrix multiplication, across different servers based on real-time performance measures such as Central Processing Unit (CPU) usage, memory utilization, time of execution, and networking latency. Model training was done with historical data obtained from past executions, incorporated into the web server to facilitate adaptive decision-making. It was tested experimentally with different levels of server scalability as well as matrix complexity. It was contrasted with a static, manual load balancer. All critical performance measures were found to be significantly improved by the AI-based methodology, with the total execution time reduced from 7,060 milliseconds to 1,000 milliseconds; network latency was also reduced to 5.12 ms, down from 214 ms; and the method reduced the overall use of CPU by 33% and overall use of memory by more than 85%. These findings confirm that intelligent, data-driven load balancing offers superior scalability, responsiveness, and efficiency for cloud-based parallel processing systems.