Comparison of Statistical and Linguistic Feature in K-Nearest Neighbors (KNN) & Neural Network Algorithms for SMS Spam Classification
Muhammad Rizal, Muhammad Fadhli, Yasir Muin · Nusantara Science and Technology Proceedings · 2025
SMS is still widely used, but the presence of spam SMS has become a serious problem. According to the 2020 Truecaller Insights Report, Indonesia recorded the highest number of spam messages in Asia, with a significant contribution from the financial services sector. This study aims to compare the influence of statistical and linguistic features in SMS spam classification using the K-Nearest Neighbors (KNN) and Neural Network (NN) algorithms. The methodology applied includes problem identification, planning, modeling, model evaluation, model implementation, and testing stages. In this research, data is processed using statistical features (TF-IDF) and linguistic features before being applied to the KNN and NN models. The performance of the models is evaluated based on precision, recall, F1-score, and accuracy metrics. The results show that the NN model using statistical features achieves an accuracy of 98%, KNN with statistical features 95%, NN with linguistic features 85%, and KNN with linguistic features 82%. Overall, the NN with statistical features outperforms the KNN in all tested feature types. From this evaluation, it can be concluded that statistical features are more effective than linguistic features, and the NN method is superior to the KNN method.