This paper introduces a novel approach to address load balancing challenges within the dynamic landscape of multi-cloud environments. The proposed scheduling model employs logistic regression to classify tasks as I/O-intensive or CPU-intensive based on their data properties, paving the way for optimized resource allocation. A multi-objective optimization technique is integrated, creating models for task completion time and cost, considering the distinct characteristics of I/O-intensive and CPU-intensive activities. The enhanced multitasking algorithm extension utilizes quadratic crossover and real number encoding to concurrently optimize multiple tasks, demonstrating superior performance in reducing computational costs, enhancing network utilization, and lowering energy consumption in a fog computing scenario. Comparative evaluations with existing algorithms, including EMTA-II, EMTA, EMT-PD, and AMT-NSGA-11, underscore the efficiency of the proposed approach in multi-cloud systems. This research contributes valuable insights into addressing the challenges of load balancing in multi-cloud environments, offering a robust solution that enhances resource allocation efficiency and scalability in cloud computing technologies.