The data generated by the IoT needs a powerful platform such as cloud computing for data processing. However, the cloud faces challenges when dealing with various types of resources, high delay, and cost, this represents a substantial challenge in scheduling tasks. Therefore, the need appeared to introduce the concept of fog. To address these limitations, optimization algorithms such as PSO were used. In traditional PSO, all particles in the swarm are influenced by a single global best particle (Gbest), if it becomes stuck in a local optimum, all the particles will move closer to it, thus, the PSO may easily get trapped in premature convergence. This paper proposed an adaptive cloud-fog integrated approach based on modified PSO called PSO Optimized Leader (PSO-OL). These modifications on four stages: Firstly, a method to ensure swarm diversity in the initialization phase is introduced. Secondly, to reduce the chance of the population getting trapped in a local optimum, the farthest-best particle is introduced. Third, in addition to the primary Gbest, a second Gbest represents a different good particle presented to explore multiple promising regions. Finally proposed a new crossover operator to get an optimized leader. The PSO-OL approach was evaluated and the results show the effectiveness of the enhanced leader by 40% with farthest-best, 45% with second-Gbest when compared to standard PSO, and when compared to scheduling algorithms where outperforms the other algorithms by minimizing makespan by 34%, cost by 14%, and increasing throughput by 75%, in comparison to existing load balancing and scheduling methods: RR, BLA, MPSO, ETS, and TCaS.