The traditional job-shop scheduling problem (JSP) requires the allocation of N independent jobs on M machines. In most flexible job-shop scheduling problem assumptions that all machines are always available. But unexpected machine failure that is called the random machine break downs is not considered. So, the stability of schedules can be computed. In this paper, an inversed model-based on random forest method in which a Gaussian process and variable importance random forest algorithm are used for mapping non-dominated solutions from the objective space (PF) to the decision space (PS). Then this proposed method is applied on an FJSP with random machine breakdowns (RMBs) which the stability of the schedule is detected by the deviation of each job time preschedule and real schedule. The proposed algorithm has been tested on the benchmark test suite for flexible job-shop scheduling problem and has compared with IM-MOEA and NSGA-II and indicates that the proposed method is a competitive and promising approach.