Apply Document Clustering to Improve Formal Concept Analysis Performance

• Main Authors: Kuan-yu Chen, 陳冠余
• Other Authors: Chuen-min Huang
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/15480580512085985812

碩士 === 國立雲林科技大學 === 資訊管理系碩士班 === 100 === Because of the problem of information overload, how to construct a good filtering mechanism for information retrieval systems is one of the most important issues. FCA is used to construct domain ontology. It can also achieve the purpose of classification in searching results based on its classification feature. However, when FCA processes large and broad datasets, it may lead to inefficient implementation by vocabulary confusion and too many attributes. This would produce enormous concept lattice and the system will spend a large amount of time while traversing concept lattice. Therefore, previously studies usually apply FCA in smaller dataset. To minimize such inefficiency, this study applies Single-pass clustering to minimize the Information dimensions before running FCA. After evaluation, we found that Single-pass clustering has a superior performance when its threshold equals 0.4. As a result, compare using FCA without Single-pass clustering, the recall improved from 3% to 37% when we extract 5% to 40% of attributes. Moreover, we found that only 10% of the attributes could reach 70% of recall. The search processing time also progressed from 4 to 15 times. In users’ satisfaction survey, 77% of the users are satisfied with our proposed method.