We address the problem of finding a set of pattern rules, from a transaction dataset given a statistical metric. A new data structure, called an incrementally counting suffix tree (ICST), is proposed for online computation of estimates of the support of any pattern or itemset. Using an ICST, our approach directly generates a set of pattern rules by a single scan of the whole dataset in partitions without the generation of frequent itemsets. Non-redundant rules can be found by removing redundancies from the pattern rules. The PPMCR algorithm first finds pattern rules and then non-redundant rules by generating valid candidates while traversing the ICST. Experimental results show that the PPMCR algorithm can be used for efficiently mining fewer non-redundant rules.