Abstract <p>A supercomputer is an expensive, in-demand resource. It is often organized as a cluster, where computational resources are managed by a scheduler and the time to obtain a result consists of two components: the time a job spends waiting in the queue and the time it takes to execute. On the one hand, a supercomputer application must efficiently utilize computational resources during execution. On the other hand, it is necessary to optimally select launch parameters, which affect both the waiting time in the queue and the execution time.</p> <p>This study proposes a lightweight approach for computational job analysis through the simultaneous consideration of multiple groups of computational jobs. Groups of jobs that are potentially of interest to a user are formed in collaboration with the user, and these groups can be gradually narrowed for further analysis.</p>