HLM-BXR: A Hybrid Learning Model Using BERT, XGBoost, and Random Forest for Spam Detection
Aragonda Vinnela, Archana Chhabra · 2025
Text classification belongs to natural language processing NLP as an important fundamental assignment that functions for spam recognition and sentiment measurement purposes. Traditional machine learning systems have difficulty capturing meaningful relationships that exist within text data. This research develops a mixed learning system based on ensemble methods and deep learning, which strengthens classification effectiveness. This model unites BERT embeddings with XGBoost and Random Forest classifiers, then uses a logistic regression meta-model for prediction enhancement. The research uses an email database where each document is labeled spam or ham. The combination of proposed hybrid model shows a high accuracy of 96.9% which proves superior to standalone classifiers. As shown by this approach deep learning features extracted by ensemble learning produce an effective and efficient classification framework for systems. The upcoming research will look into both improving the hyperparameters and applying new meta-learning strategies to boost performance rates.