A new similarity measure in formal concept analysis for case-based reasoning

摘要

In this work, we aim at developing a better knowledge base by using formal concept analysis (FCA) and propose its new similarity measure based on vector model for case-based reasoning (CBR). The features of our proposed approaches are illustrated using a part of CBR system for both classification and problem-solving. Concept lattice knowledge base provides more accuracy classification for hierarchical data structure when comparing with non-hierarchical data structure. Dependency induced from our concept lattice knowledge base can help to suggest informative solutions for problem-solving CBR. In addition, our similarity measure improves the accuracy of classification CBR significantly when we perform experiments on the UCI data sets with cross validation.