engleski

Comparison of Naive Bayes and SVM Classifiers in Categorization of Concept Maps

Concept map is a graphic tool which describes a logical structure of knowledge in the form of connected concepts. Many persons create and use concept maps as planning, knowledge representation or evaluation tool, and store them in a public repository. In such environment contents and quality of these maps vary. When user wants to use specific map, they have to know to which domain that map belongs. Many creators do not pay enough attention to complete and accurate labeling of their documents. Manually categorization of maps in large repository is almost impossible as it is a very long and demanding procedure. In such environment automatic classification of concept maps according to their content can help users to identify the relevant map. There are very few researches on automatic classification of concept maps. In this paper we propose method for automatic categorization of concept maps using simple bag of words. In our experiment, data for classification are taken from a set of public available CMs Fetched maps are filtered by language and parsed. Concepts’ labels are extracted from filtered set of CMs, preprocessed and prepared for classification. The most important features are selected and data are prepared for learning and classification. Training and classification are performed using naive Bayes and SVM classifiers. Achieved results are promising, and with further data preprocessing and adjustment of the classifiers we consider that they can be improved.

Classification; concept map; data mining; naive Bayes; SVM; text mining

nije evidentirano