Value Ordering for Solving CSP Based on Preprocessing

Preprocessing plays an important role in solving constraint satisfaction problem, as far as we know, however, all these techniques focus on reducing the scale of the problem using domain filtering. Actually, it is not the only function of them. In this paper, a value ordering searching algorithm named MAC_MPV is presented by extending preprocessing technique and exploiting those information. The proposed algorithm, during the preprocessing, not only can reduce the scale of the problem like the original technique but also can sort values of variables' domain according to the probability of the value leading to a solution. So, during searching, this ordering enforces the algorithm to search the space containing a solution as soon as possible, which means that it guides the search towards more fruitful areas of the search space. In this way, the efficiency of algorithm can be improved. Experiments on random CSPs and frb show our algorithm only requires pretty less time than conventional maintaining arc consistency.