Towards News Classification, a Machine Learning Approach Based on Stemming and Feature Extraction

News is crucial in most communities worldwide since it forms an important information source. News apps provide some privileges to keep individuals in touch with their surrounding events, such as notifications and preferences management that make users prefer news sites and apps over traditional news channels (i.e. TVs). News classification is pivotal for navigating the vast expanse of information in digital platforms. This study introduces a machine learning-based model tailored for Arabic news articles, addressing the complexities inherent in the language’s morphology. Utilising a dataset comprising 250 k articles from the National Iraqi News Agency (NINA), we applied [Formula: see text]-nearest neighbour ([Formula: see text]-NN) and logistic regression classifiers alongside feature extraction methods such as TF-IDF, Bag-of-Words and [Formula: see text]-Gram. We also utilised three stemmers (Khoja, Snowball and Tashaphyne). Our analysis revealed that while both classifiers performed commendably, the [Formula: see text]-NN classifier exhibited heightened sensitivity when integrated with [Formula: see text]-Gram features. The study’s findings underscore the potential of machine learning in advancing Arabic script-related research, with implications for information retrieval systems and digital news platforms. We rely on ([Formula: see text]1 score) metrics to analyse the model’s performance and evaluate the method performance. The evaluation is performed by analysing different classification methods, pre-processing tools and stemmers. The results showed a good classification performance for both classifiers, and there is a noticeable sensitivity for the [Formula: see text]-NN classifiers when combined with [Formula: see text]-Gram features. Also, Arabic scripts showed a remarkable complexity due to their morphological nature. The model presents outstanding results that indicate a good advancement in Arabic script-related studies.