Ensemble-Based Deep Learning Model for Identifying Malicious Text in Natural Language Processing

Abstract

 Text classification in Natural Language Processing (NLP) is crucial for sorting text data into classes. Detecting internet spam is a long-standing issue. Solutions exist for identifying spam on social media and in emails. Deep learning models are effective for text classification post-preprocessing. Preprocessing involves steps like converting text to lowercase and removing numbers, eliminating web addresses, punctuations, and stop words, as well as tokenization and stemming. Then, Feature extraction is used to generate functions from the dataset the use of area expertise and those features are extracted via Bag-of-words (BoW), word2vec, and TF-IDF. A feature set is created and it can additionally contain functions that are not meaningful or of very short duration, such capabilities want to be eliminated for better outcomes. Finally, ensemble deep learning algorithms such as Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), and Bidirectional Long Short-Term Memory (Bi-LSTM) are used to classify the malicious content within the text classification. This method is used in classifying the text and performs higher whilst compared to the existing strategies. The suggested model is implemented using the PYTHON programming language, and its performance is assessed using metrics such as accuracy, precision, recall, F-measure, balanced accuracy, brier score loss, and MAP. The suggested ensembled deep learning model achieves higher accuracy with a learning rate of 70% is 99% and a learning rate of 80% is 99.61%.