Enhanced multi-label arabic text classification based on integration of particle swarm algorithm and machine learning models

Multi-label text categorization is an important modern text mining task. The large number of feature in text datasets degrades the performance of text classification. However, multi-label text often has more noisy, irrelevant and redundant features with high dimensionality. A large amount of computational time is required to classify a large number of text documents of high dimensional. The problem is much difficult in Arabic due to complex nature of the Arabic language, which has a very rich and complicated morphology. Although a large number of studies have been proposed to other languages Multi-label text categorization, there are a few cases for Arabic multi-label data. Motivated by this, this paper proposes enhanced multi-label Arabic text classification model based of the integration of particle swarm algorithm (PSO) and three machine learning models namely Decision Tree(DT) model, k-Nearest Neighbors (KNN) model and Naive Bayes (NB) model. Experiments verify that the proposed algorithm is a useful approach of feature selection for Arabic multi-label text classification. Our experiments prove that the proposed method significantly outperforms traditional classification methods.