Optimizing Rough Set Flow Graph Inference

In this paper, we optimize Rough Set Flow Graph (RSFG) inference through several steps. Initially, we introduce the concept of barren variables and exploit them by safely removing them without necessitating any numeric computation in computer memory. Subsequently, we propose pruning independent-by-evidence variables and leveraging them in a similar manner as barren variables. Lastly, we eliminate all remaining variables by suggesting heuristics to determine effective elimination orderings for RSFG inference. Broadly, our heuristics are classified into domain heuristics and edge heuristics. Domain heuristics determine elimination orderings based on the domain cardinalities of the variables to be eliminated, while the graphical structure of RSFG with respect to edges serves as the primary factor for computing elimination orderings with edge heuristics. Our experimental analysis indicates that edge heuristics tend to produce favorable elimination orderings, whereas domain heuristics exhibit relatively less effectiveness.