A list-based compact representation for large decision tables management

Due to the huge size of the tables we manage when dealing with real decision-making problems under uncertainty, we propose turning them into minimum storage space multidimensional matrices. The process involves searching for the best order of the matrix dimensions, which is a NP-hard problem. Moreover, during the search, the computation of the new storage space that each order requires and copying the table with respect to the new order may be too time consuming or even intractable if we want a process to work in a reasonable time on an ordinary PC. In this paper, we provide efficient heuristics to solve all these problems. The optimal table includes the same knowledge as the original table, but it is compacted, which is very valuable for knowledge retrieval, learning and expert reasoning explanation purposes.