Using heuristic method to improve support vector machine for arabic text classification

Abstract EN

Text classification solves the problem of automatically assigning text content to predefined categories.

Automatic text (or document) classification attempts to replace and save human effort required in performing manual categorization.

This paper presents a performance evaluation of Support Vector Machine (SVM) in Multi Class Arabic text classification.

Introduce a heuristic method called Singular Value decomposition (SVD) to analysis the data semantically and use support vector machine to classify the data.

We compare the results with respect to SVM with heuristic method and without this heuristic method on the task of Arabic multi-class text classification.

We found that SVM with SVD substantially outperform SVM without SVD.

The results on SVM experiments show that categorization accuracy is 92.4 % and the results on SVM with SVD experiments give better accuracy 96 % for the same set.