Short Messaging Service Spam Detection Model Using Natural Language Processing and Deep Learning Techniques

• Olatunde A. AKANO

  Department of Computer Sciences, Abiola Ajimobi, Technical University, Ibadan, Oyo State, Nigeria

  Author

• Wariz A. ISMAEL

  Department of Computer Sciences, Abiola Ajimobi, Technical University, Ibadan, Oyo State, Nigeria

  Author

• Ayomikun A. AWOSEYI

  Department of Computer Sciences, Abiola Ajimobi, Technical University, Ibadan, Oyo State, Nigeria

  Author

• Femi AYO

  Department of Mathematical Sciences, Olabisi Onabanjo University, Ago-Iwoye, Ogun State, Nigeria

  Author

• Ifeoluwa M. OLANIYI

  Department of Computer Sciences, Abiola Ajimobi, Technical University, Ibadan, Oyo State, Nigeria

  Author

• Jide E.T. AKINSOLA

  Department of Computer Sciences, Abiola Ajimobi, Technical University, Ibadan, Oyo State, Nigeria

  Author

Unsolicited Short Message Service (SMS) messages, or SMS spams, pose a major challenge in mobile communication. These unwanted messages compromise user privacy, leading to data bridge or financial risks. To address this growing concern, this study explores the implementation of deep learning and Natural Language Processing (NLP) procedures to effectively detect SMS spam. By developing a robust spam detection system, this study enhances the security and usability of mobile communication platforms. This study implements an effective spam detection system using deep learning and NLP techniques. The system was developed using Python 3.10 within the Google Collaboratory environment. The SMS Spam Collection dataset, consisting of 5,574 characterized messages, underwent preprocessing procedures that included tokenization, stopword removal, lemmatization, and transformation using Term Frequency-Inverse Document Frequency (TF-IDF) vectorization. Three deep learning models were implemented for classification: Convolutional Neural Networks (CNN), Long Short-Term Memory networks (LSTM), and Recurrent Neural Networks (RNN). These models were trained and evaluated using performance metrics such as correctness, precision, recall, F1-score, Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE). Among the models tested, the CNN model demonstrated the best performance, achieving an accuracy of 96.90 percent, a precision of 0.9692, a recall of 0.9690, and an F1-score of 0.9691. It also had the lowest error rates, indicating its superior predictive capability. The results confirm the effectiveness of CNNs for SMS spam detection, particularly when combined with rigorous text preprocessing. The study suggests for further study, the application of federated leaning for modelling SMS spam detection.

• FJET_12_11_4

Copyright (c) 2025 FUDMA Journal of Engineering and Technology

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.