Neighbor knowledge construction is the foundation for the development of cooperative query answering systems capable of searching for close match or approximate answers when exact match answers are not available. This paper presents a technique for developing neighbor hierarchies at the attribute level. The proposed technique is called the evolved Pattern-based Knowledge Induction (ePKI) technique and allows construction of neighbor hierarchies for nonunique attributes based upon confidences, popularities, and clustering correlations of inferential relationships among attribute values. The technique is applicable for both categorical and numerical (discrete and continuous) attribute values. Attribute value neighbor hierarchies generated by the ePKI technique allow a cooperative query answering system to search for approximate answers by relaxing each individual query condition separately. Consequently, users can search for approximate answers even when the exact match answers do not exist in the database (i.e., searching for existing similar parts as part of the implementation of the concepts of rapid prototyping). Several experiments were conducted to assess the performance of the ePKI in constructing attribute-level neighbor hierarchies. Results indicate that the ePKI technique produces accurate neighbor hierarchies when strong inferential relationships appear among data. © 2006 IEEE.
T. Puthpongsiriporn et al., "Attribute-level Neighbor Hierarchy Construction Using Evolved Pattern-based Knowledge Induction," IEEE Transactions on Knowledge and Data Engineering, vol. 18, no. 7, pp. 917 - 929, article no. 1637418, Institute of Electrical and Electronics Engineers, Jul 2006.
The definitive version is available at https://doi.org/10.1109/TKDE.2006.104
Mechanical and Aerospace Engineering
Keywords and Phrases
Approximate query answering; Clustering; Knowledge discovery; Query-answering systems; Similarity measures
International Standard Serial Number (ISSN)
Article - Journal
© 2023 Institute of Electrical and Electronics Engineers, All rights reserved.
01 Jul 2006