<p>Cloud computing offers solutions for various scientific and business applications. Large-scale scientific applications, which are organized as scientific workflows, are carried out using cloud computing. However, the higher failure rates in cloud computing can be attributed to the numerous servers and components dealing with intense workloads. This study presents a fault-tolerant scheduling approach using proactive and clustering methods for scientific workflows in cloud computing. Initially, the task clustering issue is addressed by consolidating multiple short-duration tasks into a single job to improve the runtime performance of workflow executions. Subsequently, an automated workflow scheduling strategy is outlined with four key stages: monitoring, analysis, planning, and execution. During monitoring, clustered jobs and the capacities of available cloud resources are observed. In the analysis phase, the accuracy of failure prediction is enhanced by employing the Group Method of Data Handling (GMDH) neural network prior to any faults or failures. The planning stage introduces a novel hybrid multi-objective algorithm, MOPSO-aSA, based on MOPSO and adaptive simulated annealing (SA), to streamline workflow scheduling in error-prone execution environments. Moreover, the reliability of application execution is maintained through re-clustering and migration techniques following any faults or failures. Finally, based on the experimental findings, it is evident that the proposed strategy surpasses other methods in terms of makespan, total cost, energy consumption, and failure rate.</p>