Step Hierarchical Text Classification Based On Fuzzy Cognitive Maps

Data mining is current popular trend in world computer science field, which emerges database, artificial intelligence, machine learning, statistic, intelligent computation and cognitive science and so on. It is a novel field which can be applied in many other fields in database study, and very helpful of making decision in economic technology.Text classification is an important branch in data mining filed, which is responsible for automatically dealing with those class-unknown texts and judging which pre-defined class sets they reside in. With more and more texts put into electronical form, it is becoming more and more difficult and important for effective information retrieval and filtering. Text classification is just the efficient solution and becoming a valuable practical technology.At present, there are a large number of machine learning algorithms has been applied in text classification field, and achieving success. Class central vector method is popular and typical method. But majority of them are based on statistic computation, lack feedback system mechanism, and ignore that the whole text is an organic entity. Meanwhile, almost all of these algorithms are built on single hierarchical platform, without considering the hierarchies of text category.Thus, this paper first studies the concepts, features, knowledge representation and reasoning mechanism of Fuzzy Cognitive Maps. And then according to the feature of inner relationships and the mutual effects among nodes, as well as the whole organic entity, and fully considering the casual relationship and mutual effect among terms and categories, proposes a text classification method based on reasoning in order to classify those texts, which uses iterated reasoning mechanism, especially for the cross-subjects and new subjects which lacks of large enough texts for training and testing.Meanwhile, based on regular single hierarchical classification pattern of Chinese texts, this paper proposes the concept of text category hierarchical tree, which is based on the scale of categories. On each hierarchy, use the classification method, which combines class central vector method and reasoning algorithm based on Fuzzy Cognitive Maps, in order to construct step hierarchical classification pattern, and proposes related algorithm (namely, Step Hierarchical Text Classification). The experiment result represents that it can improve the precision and performance of text classification efficiently.