Computer Science Department, Faculty of Information Technology, Al-Ahliyya Amman University, Amman, Jordan

---

Abstract—Many Text Classification (TC) algorithms have been proposed for Arabic TC. Polynomial Neural Networks (PNNs) were used recently in English TC, and have proved to be competitive to the state of the art text classifiers in this field. Lately, they were proposed for classifying Arabic documents. In this research paper, an experimental study that directly compares PNNs against five famous classification algorithms in TC is conducted on Aljazeera-News Arabic dataset. All experiments use the same TC settings, like preprocessing, Feature Selection (FS) and reduction criteria, feature weighting and classifier performance evaluation measures. These algorithms are: SVM (Support Vector Machines), NB (Naive Bayes), kNN (k-Nearest-Neighbor), LR (Logistic Regression) and RBF (Radial Basis Function networks). Results reached in this study reveal that PNN are competitive classifiers in the field of Arabic TC.

Index Terms—Polynomial neural networks, Arabic text classification.

[PDF]

Cite: Mayy M. Al-Tahrawi, "Polynomial Neural Networks versus Other Arabic Text Classifiers," Journal of Software vol. 11, no. 4, pp. 418-430, 2016.