When publishing datasets that contain relational and transaction attributes, referred to as RT-data for briefness, either type of data may suffer from linking attacks. Anonymizing both of them is essential. However, previous approaches suffer from huge information loss during anonymizing RT-data, and they fail to preserve the utility of datasets. To address this problem, an anonymization model, (k,l)-diversity is proposed to ensure privacy by guaranteeing l-diversity on each equivalence class and k-anonymity on transaction data. In addition, two heuristic algorithms named APA and PAA, which anonymize RT-data in different orders, are also provided to achieve (k,l)-diversity. Extensive experiments based on real-world dataset show that APA and PAA outperform existing approaches in terms of execution time and information loss.