Abstract: Both traditional association rule mining and utility based association rule mining may neglect those rules whose support or utility is not high.Although these rules’ support or utility is not very high，they can satisfy those people whose main goal is to avoid risks or raise the rate of success.In order to mine the rules with a low support（or utility）and a high confidence，this paper proposes a new algorithm：HCARM.HCARM adopts partition method to handle large data，and prune out candidates by using new pruning strategy.In the meantime，by giving a proper length threshold minlen，HCARM can be fitter for long patterns mining.Experiments on synthetic data show that the method can get a good performance in mining high confidence long patterns.