Arabic Text Classification Using Learning Vector Quantization

Abstract

Text classification aims to automatically assign document in predefined category. In our research, we used a model of neural network which is called Learning Vector Quantization (LVQ) for classifying Arabic text. This model has not been addressed before in this area. The model based on Kohonen self organizing map (SOM) that is able to organize vast document collections according to textual similarities. Also, from past experiences, the model requires less training examples and much faster than other classification methods. In this research we first selected Arabic documents from different domains. Then, we selected suitable pre-processing methods such as term weighting schemes, and Arabic morphological analysis (stemming and light stemming), to prepare the data set for achieving the classification by using the selected algorithm. After that, we compared the results obtained from different LVQ improvement version (LVQ2.1, LVQ3, OLVQ1 and OLVQ3). Finally, we compared our work with other most known classification algorithms; decision tree (DT), K Nearest Neighbors (KNN) and Naïve Bayes. The results presented that the LVQ's algorithms especially LVQ2.1 algorithm achieved high accuracy and less time rather than others classification algorithms and other neural networks algorithms.