<p>The Internet of Things (IoT) is an emerging technology, increasingly integrated into diverse everyday devices. It promises to enhance human lives by improving accessibility of information and delivering efficient and high-quality services. IoT devices communicate with each other over the Internet to create more complex applications and services. This has led to the IoT service composition problem, where multiple services with equivalent functionalities but different quality of service (QoS) attributes must be selected and combined to meet user needs. In large-scale environments, finding an optimal combination of services that satisfies global QoS constraints is an NP-hard problem, particularly when considering the trade-off between optimizing QoS and ensuring feasibility of the composition. Various approximation approaches have been proposed to tackle the QoS-aware service composition problem under global constraints. While many of these approaches rely on meta-heuristics to explore solution spaces, they often overlook the possibility that the optimal solution might lie on the boundary between feasible and infeasible regions. Additionally, some approaches failed to consider the trade-off between optimizing the objective function and satisfying the user constraints. This paper introduces MVCH-EFWA, a QoS-aware IoT services composition approach that employs the Enhanced Fireworks Algorithm (EFWA) to efficiently discover near-optimal compositions while minimizing computational time. To effectively handle user constraints, the approach incorporates the Modified Violation Constraints Handling technique (MVCH). MVCH-EFWA excels at effectively navigating the boundary between feasible and infeasible regions to identify optimal solutions. The results obtained through the simulation scenarios show that MVCH-EFWA yields good results in terms of utility, execution time, and degree of violated constraints.</p>