A Context-Aware Personalized Recommendation Framework Integrating User Clustering and BERT-Based Sentiment Analysis

Abstract

With the rapid growth of e-commerce platforms, there is an increasing demand for highly accurate and personalized recommendation systems. Traditional recommendation algorithms often struggle to capture the complex and dynamic nature of user preferences, especially when dealing with heterogeneous data sources. This study introduces a novel recommendation framework that integrates user clustering, BERT-based sentiment analysis, contextual encoding, and deep learning techniques. Utilizing a real-world dataset from Kaggle, the proposed model incorporates user behavior records, review texts, and contextual information to construct detailed user and item representations. Dimensionality reduction and clustering methods are employed to identify latent user groups, while BERT is used to extract deep semantic features from user-generated reviews. The resulting feature vectors are then fed into a multi-layer perceptron to generate personalized recommendations. Extensive experiments demonstrate that the K-means + BERT + MLP model consistently outperforms a variety of traditional and hybrid baselines across multiple evaluation metrics, including accuracy, precision, recall, F1-score, and AUC. The results validate the effectiveness and robustness of the proposed approach, showcasing the potential of multi-source feature fusion and advanced modeling techniques in next-generation recommender systems.