Veri Madenciliǧi Teknikleriyle Türkçe Web Sayfalarinin Kategorize Edilmesi

Today, it is not possible to use human power alone to cope with the increasing amount of data. For this reason, some automated methods are needed to group similar documents together or to place documents in predefined categories according to certain rules. The use of automated classification techniques is becoming increasingly important for this reason. In this study, a database consisting of 22 thousand samples was created in order to respond to the need for Turkish data and various methods used for text classification in the literature were tested on this data. Multinomial Naive Bayes (M-NB) and Support Vector Machines (SVM) algorithms which are frequently used for text classification, were compared by applying the n-gram word vector selection and information gain ratio. Beside these, it has been focused on the number of categories, the content of data used to train the model and the completeness of this data, and also the effects of these on classification success are examined.