A Comparison of Different Kernel Functions of SVM Classification Method for Spam Detection

Abstract

Today, the use of e-mail, especially for formal online communication, is still often done. There is one common problem faced by e-mail users, which is the frequent receiving of spam messages. Spam messages are generally in the form of advertising or promotional messages in bulk to everyone. Of course this will cause inconvenience for people who receive the SPAM message. SPAM e-mails can be interpreted as junk messages or junk mail. So that spam has the nature of sending electronic messages repeatedly to the owner of the e-mail. This is abuse of the messaging system. One way to solve the spam problem is to identify spam messages for automatic message filtering. Several machine learning based methods are used to classify spam messages. In this study, a comparison was made between several kernel functions (i.e., linear, degree 1 polynomial, degree 2 polynomial, degree 3 polynomial, and RBF) of the SVM method to get the best SVM model in identifying spam messages. The evaluation results based on the Kaggle 1100 dataset showed that the best model were the SVM model with a linear kernel function and a degree 1 polynomial, where both models returned Precision = 0.99, Recall = 0.99, and F1-Score = 0.98. On the other hand, the RBF kernel produced lower performance in terms of Precision, Recall, and F1-Score of 0.95, 0.95, and 0.94, respectively.

Spam Email Detection Using Machine Learning Techniques

“…Since identification is a specific task of classification, machine learning is usually used to classify email into two classes, namely spam and benign messages, also called "ham". Several studies using machine learning-based spam classification have been published in the research bibliography (Ghosh & Senthilrajan, 2023;Karyawati et al, 2023). Previous research work on anti-spam filtering studied the performance of popular machine learning algorithms (Michelakis et al, 2004;Ahmed et al, 2022;Ghosh & Senthilrajan, 2023).…”

“…However, it requires large and unbiased amounts of data. In addition, parameter adjustment is required to select the best model (Karyawati et al, 2023). Laorden (2012) developed a Word Sense Disambiguation preprocessing step before applying machine learning algorithms to detect spam data.…”

“…Nonlinear kernel functions are often used to convert the input data to a high-dimensional feature space where the input data becomes more separable than the original input space (Karyawati et al, 2023).…”

Enhancing Fault Detection in Wireless Sensor Networks Through Support Vector Machines: A Comprehensive Study

“…The problem is to determine a hyperplane in which the space 𝑂 can be divided linearly by solving the following minimization problem (2): min: 𝐾 = (3) subject to ∑ 𝑙 𝑖=1 𝜉 𝑖 𝑦 𝑖 = 0, 𝐶 ≥ 𝜉 𝑖 ≥ 0, where parameter 𝐶 is a parameter that is used to control between the margin and the learning error. 𝐹(𝑥 𝑖 ⋅ 𝑥 𝑗 ) -a Kernel Function [73] designed to transform input data into a high-dimensional feature space is required to implement SVM. As a result, the nonlinear SVM function is described as (4):…”

Towards Optimal Solar Energy Integration: A Deep Dive into AI-Enhanced Solar Irradiance Forecasting Models

“…This step ensures that our data is consistently normalized, a prerequisite for algorithms like SVM to function optimally. We harness the power of the SVM for our predictions, conducting a comprehensive search over specific kernel parameters, namely linear and poly [80,81]. Our meticulous exploration leads to the linear kernel as the superior choice, a testament to the efficacy of our methodological approach.…”