Effetcts of dimensionality reduction and feature selection in text categorization

The goal of classifying text or generally data is to decrease the time of access to the information. Continuously increasing of documents makes the classification process impossible to do manually. In this case, the automatic text classification systems are activated. In text classification systems, large data space is an important problem. By using dimensionality reduction techniques and feature selection in text classification systems, it is possible to do right classification with reduced size of data. In this study, all of the texts have been represented by term frequency ? inverse document frequency (TF?IDF) vectors. Discrete Cosine Transform (DCT) method and the feature selection with Proportion of Variance method are used to get more effective results for classification by reducing dimensionality of TF?IDF vector space which is generated by text vectors. Successful results have been obtained with reduced dimension. In addition, these results are obtained in a short time. Milliyet, R8 and WebKB?4 datasets are used. Milliyet dataset contains 5 classes and is prepared by us. R8 dataset is in Reuters?21578, contains eight classes and is used in literature. WebKB?4 dataset is collected from web sites of computer science departments of various universities, contains four classes and is used in literature. For all three datasets, texts are grouped in training and test texts. Classification system is trained by training test and classes of test texts are determined by using our system. According to chosen method, classification success increases 92% with reduced dimension of vector.Proposed methods are applied with using Microsoft .Net framework and C# language.