基于粒度计算的减量式隐私保护数据挖掘

摘要

In order to work better on privacy preserving data mining,described the research progress in this area.Aiming at EMASK algorithm-the typical method based on random perturbation,this paper proposed improvements with granular compu-ting,transforming the relational data forms into granularity relation model for machine.With bit operation method to ensure no reduction of accuracy,this calculation method used vertical Bitmap representation of the data,reducing the number of I /O op-erations and the space overhead.At the same time,it also recorded the distorted data support and reduced file access times in the generation of frequent item sets,thus improved the calculation efficiency.In view of the real world database changes,it solved the calculation problem of decreasing business database frequent item sets by using the decreasing updating algorithm technology.The experimental results show that,whether in fixed or variable decreasing database processing,the time efficien-cy improves greatly by BDEMASK compared with EMASK.%为了能够更好地开展隐私保护数据挖掘工作，描述该领域的研究进展。针对基于随机的干扰方法中典型代表 EMASK 算法，用粒度计算的思想提出改进意见，将关系数据表转换成面向机器的粒度关系模型。这种计算方法使用了数据的垂直 Bitmap 表示，利用位操作的方法来保证准确性不降低的情况下，减少 I ／O 操作的次数，降低空间开销，同时在生成频繁项集时，也记录了其在扭曲后数据中的支持度，减少了文件的访问次数，由此提高计算效率。针对现实世界事务数据库变化情况，利用减量式更新算法技术来解决减量式事务数据库频繁项集计算问题。实验结果证明，无论是在固定减量集数据库还是可变减量集数据库处理中，BDEMASK 相对于EMASK 而言，时间效率都有很大幅度的提高。