Integer Bloom Filter: Achieving Low-Error Multi-Attribute Membership Querying

• Main Authors: HUNG-YU CHENG, 鄭弘裕
• Other Authors: HENG MA
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/d3fznv

博士 === 中華大學 === 科技管理博士學位學程 === 102 === This thesis proposes Integer Bloom Filter for multi-attribute membership querying of identifiers, which combines the concepts of Bloom Filters and artificial neutral networks. Membership querying of identifiers has been applied on a wide range of categories, and normally this type of problem could be dealt with by employing traditional binary-based Bloom Filters. Recently, as the data amount on the internet has been dramatically increased, multi-attribute membership querying has started to be emphasized by researchers because computational effectiveness can be enhanced with such a technique. Traditional Bloom Filters with multi- layer or segment implementations are capable of solving this problem; however, such implementations could result in lower computational effectiveness and higher error rates. In this thesis, a novel type of Bloom Filter with integer data is developed in an attempt to resolve the problems stated above, where the training process of artificial neural networks is incorporated. The proposed training algorithm is associated with both insertion and deletion operations, which is critical in achieving the goals of this thesis. Two experimental data were simulated including (1) car license plates in Taiwan and (2) email accounts. The results show, with appropriate parameter settings, the error rates could be controlled under a satisfactory level with intact computational effectiveness. Future research will be focused on non-string data patterns including images and voice signals.