The exponential rise in online activity has increased the possibility of phishing attempts, which can seriously jeopardize sensitive data and put companies and individuals at risk. This study uses cutting-edge machine learning and deep learning approaches to solve the urgent need to identify phishing websites efficiently. The proposed research utilizes two different extensive datasets with 30 and 87 features, respectively, that have been extracted from various network traffic, site content, and URLs to broadly measure the binary classification performance and accuracy of 5 different machine learning models as well as 3 deep learning-based models. Furthermore, this research employed one of the most popular feature selection techniques, Pearson correlation, to reduce the number of input variables, minimizing the amount of redundant or irrelevant features in the datasets. Training our machine learning and deep learning models with the resulting mostly redundancy-free subset of features was very efficient and with the addition of hyperparameter tuning and 10-fold cross-validation, our models produced some of the most accurate prediction results in web phishing detection. Among all the tests, the Random Forest algorithm performed better in the limited-sized dataset and scored a prediction accuracy of 97.83%. On the larger dataset, eXtreme Gradient Boosting performed slightly better and achieved a prediction accuracy of 97.02%.