Modified Genetic Algorithm for Feature Selection and Hyper Parameter Optimization: Case of XGBoost in Spam Prediction

Recently, spam on online social networks has attracted attention in the research and business world. Twitter become preferred medium to spread content. Many efforts attempted encounter spam. brought extra challenges represented by feature space size, imbalanced data distributions. Usually, related works focus part of these main or produce black-box models. In this paper, we propose a modified genetic algorithm for simultaneous dimensionality reduction hyper parameter optimization over datasets. The initialized an eXtreme Gradient Boosting classifier reduced features tweets dataset; generate prediction model. model is validated using 50 times repeated 10-fold stratified cross-validation, analyzed nonparametric statistical tests. resulted attains average 82.32% 92.67% terms geometric mean accuracy respectively, utilizing less than 10% total space. empirical results show that outperforms Chi2 PCA selection methods. addition, many machine learning algorithms, including BERT-based deep model, prediction. Furthermore, proposed approach applied SMS modeling compared works.