Text categorization with machine learning techniques

The aim of this thesis is to study the mathematical model of text classification algorithms, being one of the most important problems of present time, and then, to apply it to the real life by using the context of text mining which is a sub branch of data mining. There have been a lot of studies in literature about this subject. In this thesis, these previous workings will be studied while their results will be interpreted. In the frame of this thesis, firstly, text mining and the problem of text classification will be introduced. After that, the formulation of classification problem will be explained. Before applying the text classification problem, the pre-calculation processes of text data will be clarified to use in the problem. Following these pre-calculation processes; being aimed at vector-space model, at taking information from text and at increasing the percent of correct classifying, the feature selection algorithms will be discussed. Besides, developed for text classification, the classification algorithms will be illustrated and then compared with each other by presenting their positive and negative sides. In the last section, there are the numerical results of algorithm performances. The influence of feature selection algorithms on the classification percent will be argued by using obtained results. Final results will be interpreted with the help of graphs and presented in tableaus. Keywords: Text Classification, Vector-Space Model, Naive Bayes, Multinomial Naive Bayes, Complement Naive Bayes, K-Nearest-Neighbor, Information Gain, Gain Ratio, Chi- Square