SMS DATASET FOR MULTI-CLASS CLASSIFICATION OF HAM, SPAM, AND SMISHING IN AZERBAIJANI LANGUAGE

The developing landscape of digital communication necessitates specialized datasets for effective research and development, particularly in under-resourced languages. This study introduces Azerbaijani SMS Collection, a novel multi-class dataset specifically designed for the classification of Azerbaijani SMS messages into legitimate, promotional, and malicious phishing categories. To demonstrate the utility of Azerbaijani SMS Collection, an extensive evaluation was conducted using a broad spectrum of machine learning and deep learning models. Traditional machine learning methods, including Logistic Regression, Linear SVM, Passive Aggressive Classifier, Multinomial Naive Bayes, Decision Tree, Random Forest, and K-Nearest Neighbors, were employed in the assessment. The dataset was further evaluated using deep learning architectures, specifically Convolutional Neural Networks and Recurrent Neural Networks. Among all evaluated models, Convolutional Neural Networks achieved the strongest overall performance, with an accuracy of 0.9393 and an F1-score of 0.8909, while the Passive Aggressive classifier was the best-performing traditional algorithm with an accuracy of 0.9338 and an F1-score of 0.8821. This research provides a valuable foundation for future studies in Azerbaijani text classification and contributes significantly to efforts aimed at enhancing SMS communication security and combating digital fraud in the region (pp.32-39).