Title: A new feature selection algorithm for evolutionary analysis of Aramaic and Arabic script variants

Authors: Osama A. Salman; Gábor Hosszú; Ferenc Kovács

Addresses: Faculty of Electrical Engineering and Informatics, Department of Electron Devices, Budapest University of Technology and Economics, 1111 Budapest, Műegyetem rkp. 3., Hungary ' Faculty of Electrical Engineering and Informatics, Department of Electron Devices, Budapest University of Technology and Economics, 1111 Budapest, Műegyetem rkp. 3., Hungary ' Faculty of Information Technology and Bionics, Pázmány Péter Catholic University, 1083 Budapest, Práter u. 50/A, Hungary

Abstract: This paper deals with applying phylogenetic modelling to the evolution of scripts (writing systems) as taxa. Aramaic and Arabic script variants are studied in the present cladistic analysis. The selection of the most suitable features of taxa for accurate modelling as part of the feature engineering step could improve the result of the cladistic analysis. The main objective is to filter out features of the taxa under study that could potentially cause homoplasy. The effect of feature filtering is investigated using some widely used phylogenetic software products for biological databases. Studies have consistently shown that the phylogenetic tree (cladogram) generated after filtering out the most variable features is more optimal for less homoplasy than the tree obtained without feature filtering. Hence, the proposed algorithm effectively pre-filters the features that may cause homoplasy. Furthermore, the results also demonstrated that different cladistic methods investigated gave similar results for the dataset under study.

Keywords: Arabic script; Aramaic script; cladistics; evolutionary analysis; feature selection; maximum likelihood; maximum parsimony; pattern evolution; pattern system; scriptinformatics.

DOI: 10.1504/IJIEI.2022.128892

International Journal of Intelligent Engineering Informatics, 2022 Vol.10 No.4, pp.313 - 331

Received: 11 May 2022
Accepted: 16 Sep 2022
Published online: 08 Feb 2023 *

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