A Method for SMS Spam Message Detection Using Machine Learning

Abstract

In recent years, it has become increasingly common for individuals to connect with their relatives and friends, read the most recent news, and discuss various social topics by using online social platforms such as Twitter and Facebook. As a consequence of this, anything that is considered spam can quickly spread among them. The identification of spam is widely regarded as one of the most significant problems involved in text analysis. Previous studies on the detection of spam concentrated primarily on English-language content and paid little attention to other languages. The information gathered by the University of California; Irvine served as the basis for the development of our spam detection technology (UCI). In this research study, the effectiveness of various supervised machine learning algorithms, such as the J48, KNN, and DT, in identifying spam and ham communications is investigated. SMS spam is becoming more widespread as the number of individuals who use the internet continues to rise and an increasing number of businesses disclose their customers' personal information. E-mail spam filtering is the progenitor of SMS spam filtering, which inherits a significant number of its features. We evaluate the proposed method based on accuracy, recall, and precision. Experiments showed that DT has obtained higher accuracy compared to other machine learning classifiers.

Detection of Korean Phishing Messages Using Biased Discriminant Analysis under Extreme Class Imbalance Problem

“…In [12], the performance of phishing detection was compared using machine learning algorithms kNN, and DT. The dataset consisted of 747 phishing data and 4827 non-phishing data.…”

“…A range of machine learning algorithms have been applied to spam detection [8][9][10][11][12][13][14][15][16][16][17][18][18][19][20][21]23,24,34]. These include traditional methods such as SGD, SVM, and NB, alongside deep learning-based techniques like ANN, CNN, LSTM, BiLSTM, and GRU.…”

“…Moreover, non-parametric supervised learning algorithms such as DT and RF and the clustering method k-means have been utilized. The target languages for spam detection research include English [8][9][10][12][13][14][15]17,[19][20][21]23,34], Indonesian [18], Bengali [16], Arabic [24], and Turkish [11]. These studies converted text data into embedding vectors using techniques such as TF-IDF, BoW, Bidirectional Encoder Representations from Transformers (BERT), and Continuous Bag of Words (CBoW), making them suitable for training.…”

“…In the majority of the previous studies, various machine learning algorithms were employed to improve the performance of phishing messages detection in imbalanced datasets [8][9][10][12][13][14][15]17,21,23,34].…”

“…Traditional detection algorithms include Naive Bayes (NB), Support Vector Machine (SVM), and Logistic Regression (LR) [8][9][10][11]. Moreover, tree-based methods are represented by Decision Tree (DT) and Random Forest (RF), while boosting methods include Adaptive Boosting (AdaBoost), Extreme Gradient Boosting (XGBoost), and Light Gradient Boosting Model (LGBM) [12][13][14][15][16][17]. Additionally, detection techniques using neural networks involve Stochastic Gradient Descent (SGD), Artificial Neural Network (ANN), Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM), with the inclusion of Bidirectional Long Short-Term Memory (BiLSTM) to learn bidirectional long-term dependencies by adding backward LSTM [18][19][20][21].…”