In recent years, privacy preserving data publishing attracted many attentions due to the concern of privacy breaches. The removing of personal identifiable information, such as the naïve anonymization, is not sufficient. The privacy preserving data publishing technologies transform data into a form that sensitive personal information cannot be identified and retaining to the greatest extent possible usefulness of published data. In this work, we propose a novel strategy to deal with the sensitive items and quasi-identifier items separately. The proposed algorithm has at least the same or stronger privacy level for k-anonymity on transactional data, 1/ k. According to the numerical experiment results, our proposed strategy has better performance on running time, better data utility.