Recent advancements in applications such as natural language processing (NLP), applied linguistics, indexing, data mining, information retrieval, and machine translation have emphasized the need for robust datasets and corpora. While there exist many Arabic corpora, most are derived from social media platforms like X or news sources, leaving a significant gap in datasets tailored to academic research. To address this gap, the ARPD, Arabic Research Papers Dataset, is developed as a specialized resource for Arabic academic research papers. This paper explains the methodology used to construct the dataset, which consists of seven classes and is publicly available in several formats to benefit Arabic research. Experiments conducted on the ARPD dataset demonstrate its performance in classification and clustering tasks. The results show that most of the classical clustering algorithms achieve low performance compared to bio-inspiration algorithms such as Particle Swarm Optimization (PSO) and Gray Wolf Optimization (GWO) based on the Davies–Bouldin index measure. For classification, the Support Vector Machine (SVM) algorithm outperformed others, achieving the highest accuracy, with other classifiers ranging from 89% to 99%. These findings highlight the ARPD’s potential to enhance Arabic academic research and support advanced NLP applications.