<![CDATA[The increasing use of Short Message Service (SMS) in digital communication has been accompanied by a rise in spam messages, which threaten user convenience and information security. This study presents a comparative analysis of three classical machine learning algorithms—Decision Tree, Naïve Bayes, and Logistic Regression—for SMS spam classification. The research follows the CRISP-DM methodology, including data collection, understanding, preparation, modeling, and evaluation. The dataset used is the SMS Spam Collection (A More Diverse Dataset) from Kaggle, comprising 5,574 SMS messages labeled as spam or ham. Text preprocessing is performed through cleaning operations and feature extraction using the Term Frequency–Inverse Document Frequency (TF-IDF) method. The models are evaluated using accuracy, precision, recall, F1-score, and Area Under the Curve (AUC) metrics. Experimental results indicate that Logistic Regression achieves the most balanced performance, with an accuracy of 97.13%, precision of 99.23%, recall of 80.75%, F1-score of 89.04%, and an AUC of 98.72%. Naïve Bayes demonstrates high efficiency and perfect precision but lower recall, while Decision Tree offers interpretability with comparatively lower classification performance. The results suggest that Logistic Regression is the most suitable model for lightweight and reliable SMS spam detection systems, balancing accuracy and misclassification risk. This study provides practical insights for implementing efficient spam filtering solutions and serves as a reference for future research in text classification and natural language processing, particularly for short-message communication.]]>
Copyrights © 2025
