Phishing attacks are on the rise in current society where attackers try to steal sensitive information from the user using any means necessary. Majority of the attacks leverage electronic mails to trick unsuspecting users to click a link or urge a sense of urgency. Thus this research work aims to explore key features of phishing emails for the detection of the phishing attacks. The main focus was to apply Exploratory Data Analysis (EDA) on key features like URL Patterns, timestamps, sentiment of sender, grammar and TF-IDF (Term Frequency-Inverse Document Frequency) which can be leveraged in detection of phishing attacks. Using these features, machine learning models like Random Forest (RF), XG Boost (eXtreme Gradient Boosting) were evaluated on the dataset mentioned in Sect. 3.1 and it was found that XG Boost provided the False Positive Rate of just 1.76% with the lowest execution time of just 27.85 s. In comparison, the best performing Bidirectional Long Short-Term Memory (Bi-LSTM) model had a False Positive Rate of 1.20% and had a very high execution time. The results highlight XGBoost as the optimal choice for balancing performance and computational efficiency for the detection of phishing emails.