Spam and Phishing emails are the most crucial in social networks, many issues arise through emails such as cost of dealing with spam and phishing emails due to their large quantities, privacy resulting in loss of sensitive information, time taken to identify spam and phishing emails, and cyber security threat due to malicious content. Using a spam and phishing detection approach, a model can quickly recognize spam and phishing emails and classify them before they become a threat to the organization. In this study, a machine learning and Natural Language processing-based supervised learning approach was used and plays an effective role in improving email classification. The dataset was prepared and dynamically classified into 3 categories namely spam-ham, spam-phishing, and ham-phishing. Different methods for effective classification were performed such as data preprocessing, feature selection, model training, model testing, and classification result and performance evaluation. There were 5 machine learning algorithms used, and the result was evaluated using 8 performance indexes. The result shows that the XGBoost classifier out-performed other machine learning algorithms used, Results show that XGBoost machine learning algorithms outperformed other algorithms using the datasets. This research would help to improve categorizing emails into different folders based on their content, intent, or relevance, improve user experience, and better manage email inboxes by automatically filtering, sorting, and prioritizing messages.