Proceedings of the International Workshop "What can FCA do for Artificial Intelligence?" (FCA4AI at ECAI 2016)
The four preceding editions of the FCA4AI Workshop showed that many researchers working in Artificial Intelligence are deeply interested by a well-founded method for classi- fication and mining such as Formal Concept Analysis (see http://www.fca4ai.hse.ru/). The first edition of FCA4AI was co-located with ECAI 2012 in Montpellier, the second one with IJCAI 2013 in Beijing, the third one with ECAI 2014 in Prague, and finally the forth and last one with IJCAI 2015 in Buenos Aires. In addition, all the proceedings of these pre- ceding editions have been published as CEUR Proceedings (http://ceur-ws.org/Vol-939/, http://ceur-ws.org/Vol-1058/, http://ceur-ws.org/Vol-1257/ and http://ceur-ws. org/Vol-1430/).
This year, the fifth workshop has again attracted many different researchers working on actual and important topics, e.g. theory, fuzzy FCA, dependencies, classification, mining of linked data, navigation, visualization, and various applications. This shows the diversity and the richness of the relations between FCA and AI.
Formal Concept Analysis (FCA) is a mathematically well-founded theory aimed at data analysis and classification. FCA allows one to build a concept lattice and a system of de- pendencies (implications) which can be used for many AI needs, e.g. knowledge discovery, learning, knowledge representation, reasoning, ontology engineering, as well as information retrieval and text processing. As we can see, there are many “natural links” between FCA and AI. Recent years have been witnessing increased scientific activity around FCA, in particular a strand of work emerged that is aimed at extending the possibilities of FCA w.r.t. knowl- edge processing, such as work on pattern structures and relational context analysis. These extensions are aimed at allowing FCA to deal with more complex than just binary data, both from the data analysis and knowledge discovery points of view and as well from the knowledge representation point of view, including, e.g., ontology engineering. All these in- vestigations provide new possibilities for AI activities in the framework of FCA. Accordingly, in this workshop, we are interested in two main issues:
How can FCA support AI activities such as knowledge processing (knowledge discov- ery, knowledge representation and reasoning), learning (clustering, pattern and data mining), natural language processing, and information retrieval.
How can FCA be extended in order to help AI researchers to solve new and complex problems in their domains.
The workshop is dedicated to discuss such issues. This year, the papers submitted to the workshop were carefully peer-reviewed by three members of the program committee and 14 papers with the highest scores were selected. We thank all the PC members for their reviews and all the authors for their contributions.
Decision tree learning is one of the most popular classifica- tion techniques. However, by its nature it is a greedy approach to finding a classification hypothesis that optimizes some information-based crite- rion. It is very fast but may lead to finding suboptimal classification hy- potheses. Moreover, in spite of decision trees being easily interpretable, ensembles of trees (random forests and gradient-boosted trees) are not, which is crucial in some domains, like medical diagnostics or bank credit scoring. In case of such “small, but important-data” problems one is not obliged to perform a greedy search for classification hypotheses, and therefore alternatives to decision tree learning techniques may be con- sidered. In this paper, we propose an FCA-based classification technique where each test instance is classified with a set of the best (in terms of some information-based criterion) classification rules. In a set of bench- marking experiments, the proposed strategy is compared with decision tree and nearest neighbor learning.