Spam Detection using NLP and Machine Learning Techniques

Spam is one of the most common cyber-attacks in today’s digital landscape. This paper compares spam detection capabilities of classical Machine Learning models using Inverse Document Frequency text processing with Deep Learning models using Corpus Indexing text processing using NLP and shows that Deep Learning models outperform classical ML models. As a result, SVM achieved 98.11% accuracy amongst the classical ML models whereas LSTM achieved an exceptional 98.85% accuracy. The BERT model achieved 99.29% accuracy, the highest for this use case. This paper discusses implementation, compares the various results from different authors and shows how Deep Learning helps to achieve the highest levels of accuracies possible for this use case.