Title: Combining RSS-SVM with genetic algorithm for Arabic opinions analysis

Authors: Amel Ziani; Nabiha Azizi; Djamel Zenakhra; Soraya Cheriguene; Monther Aldwairi

Addresses: Lri Laboratory: Computer Research Laboratory, Badji Mokhtar University, Annaba, 23000, Algeria ' Labged Laboratory: Electronic Documents Control Laboratory, Badji Mokhtar University, Annaba, 23000, Algeria ' Labged Laboratory: Electronic Documents Control Laboratory, Badji Mokhtar University, Annaba, 23000, Algeria ' Labged Laboratory: Electronic Documents Control Laboratory, Badji Mokhtar University, Annaba, 23000, Algeria ' College of Technological Innovation, Zayed University, P.O. Box 144534 Abu Dhabi, UAE

Abstract: Due to the large-scale users of the Arabic language, researchers are drawn to the Arabic sentiment analysis and precisely the classification areas. Thus, the most accurate classification technique used in this area is the support vector machine (SVM) classifier. This last, is able to increase the rates in opinion mining but with use of very small number of features. Hence, reducing feature's vector can alternate the system performance by deleting some pertinent ones. To overcome these two constraints, our idea is to use random sub space (RSS) algorithm to generate several features vectors with limited size; and to replace the decision tree base classifier of RSS with SVM. Later, another proposition was implemented in order to enhance the previous algorithm by using the genetic algorithm as subset features generator based on correlation criteria to eliminate the random choice used by RSS and to prevent the use of incoherent features subsets.

Keywords: Arabic opinion mining; SentiWordNet; machine learning; SVM; support vector machine; RSS; random sub space; GA; genetic algorithm.

DOI: 10.1504/IJISTA.2019.097754

International Journal of Intelligent Systems Technologies and Applications, 2019 Vol.18 No.1/2, pp.152 - 178

Received: 23 Feb 2017
Accepted: 27 Aug 2017
Published online: 07 Feb 2019 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article