Spam Feature Selection Using Harris Hawks Optimization Algorithm
Adeeb M. Alsaaidah, Hani Almimi, Ahmad Adel Abu-Shareha, Mosleh M. Abualhaj, Mahran Al Zyoud · 2025
The identification of spam is crucial for networks and cybersecurity. The Internet has historically served as a line for cybercriminal activities such as spam. This study employs supervised Machine Learning (ML) to identify and detect spam. The ISCX-URL-2016 dataset from the Canadian Institute for Cyber Security is utilized for assessment purposes. This dataset has 79 features and 14,478 samples, categorized as spam and benign. The Python tools are used to evaluate and train the ML methods. The Harris Hawks Optimization (HHO) approach is utilized for dataset dimensionality reduction, with critical features selected according to their importance. The top 10 features were identified using the HHO method, and the classifiers were trained to employ 5-fold cross-validation. The ML methods assessed are Logistic Regression (LR) and K-Nearest Neighbors (KNN). These two ML methods have been tailored using the Random Serach (RS) method to address the spam detection issue. The performance parameters of accuracy, precision, and recall are evaluated. The KNN classifier achieved the best accuracy of 99.59%.