Classification of SPAM mail utilizing machine learning and deep learning techniques

Abstract: The Internet and social media networks usage has increased nowadays and become a prominent medium of communicating. Email is one of the professional reliable methods of communication. Automatic classifications of spam emails have become an area of interest. In order to detect spam emails, this study utilizes a dataset, including spam and non-spam emails. Various techniques are applied to obtain higher accuracy using machine learning techniques. NLP is also utilized for improvising accuracy using embeddings. For that, this work utilizes the BERT model, to achieve satisfactory detection of spam emails. Further, the results are compared with state-of-the-art methods, including, KNN, LSTM and Bi-LSTM. The results obtained by Bi-LSTM and LSTM were 97.94% and 86.02%, respectively. The presented methodology is promising in detecting spam emails due to the higher accuracy achieved.