Improved Spam Detection Through LSTM- Based Approach

In the era of evolving digital environment, users are constantly plagued by unsolicited e-mail and SMS, which calls for innovative solutions to protect digital communications. This research seeks to solve this ongoing hurdle by harnessing the strength of Deep Learning (DL) models, specifically using Natural Language Processing (NLP) techniques and Long-Short-Term Memory (LSTM) networks. The objective of proposed study is to find efficient detection algorithm for e-mails and SMS. The study embarks on a complex journey of text data preprocessing that includes character cleaning, case uniformity, lemmatization, and stop word elimination. Its highlight is the use of a deep learning model built on LSTM for spam classification. The processed text is then encoded using one fast encoder, and an LSTM-based deep learning model is applied to label spam, regardless of the message source. The model contains basic layers, including embedding layer, bidirectional LSTM layer, and sigmoid-activated dense layer driven by binary cross-entropy loss and Adam Optimizer. The results achieved remarkable improvements that outperform traditional rule-based filters and substantially reduce false positives in email and SMS classification. The research highlights the potential of NLP and LSTM-based models in revolutionizing spam detection, making it more accurate providing the accuracy of 96.19% which is more promising.