International Journal of Innovative Research in Computer and Communication Engineering
ISSN Approved Journal | Impact factor: 8.771 | ESTD: 2013 | Follows UGC CARE Journal Norms and Guidelines
| Monthly, Peer-Reviewed, Refereed, Scholarly, Multidisciplinary and Open Access Journal | High Impact Factor 8.771 (Calculated by Google Scholar and Semantic Scholar | AI-Powered Research Tool | Indexing in all Major Database & Metadata, Citation Generator | Digital Object Identifier (DOI) |
| TITLE | Ensemble Deep Learning Approach for Drug Effectiveness Prediction using Natural Language Processing |
|---|---|
| ABSTRACT | In today’s world, the growing abundance of reviews from patients about drugs in real-time gives us a unique opportunity to enhance the evaluation of the effectiveness of drugs in practice beyond clinical research. However, while this possibility exists, it is daunting to use user-generated content from unstructured medical text, and the challenge is com pounded by the variability in language and the variability of personal response. In this paper, we present a hybrid machine learning framework to classify drug effectiveness, using more than 2,15,000 patient reviews and publicly posting their thoughts and experiences online. The proposed model is based on natural language processing methods and ensemble learning to forecast patient’s perceived drug effectiveness from structured and un structured data, including drug name, disease, review text, rating, and number of helpful votes. We process the review data into term frequency-inverse document frequency (TF-IDF) vectors which we then use as features for a number of well-known classifiers (Logistic Regression, Random Forest, and XGBoost), as well as for a bidirectional LSTM deep learning model to capture deeper semantics of the text within the context. Outputs from these separate classifiers are combined through a stacked ensemble meta-classifier to improve generalization and robust ness. Evaluation of the framework on stratified default datasets demonstrates a further improvement in F1-score and ROC-AUC over the solo classifiers and AutoML benchmarks. Taken together, these findings show that combining contextual deep learning and structured features represents a robust, interpretable and scalable method for drug review analytics and pharmacovigilance. |
| AUTHOR | K ASHA DEEPTHI, B MADHAV RAO |
| VOLUME | 176 |
| DOI | DOI: 10.15680/IJIRCCE.2025.1311061 |
| pdf/61_Ensemble Deep Learning Approach for Drug Effectiveness Prediction using Natural Language Processing.pdf | |
| KEYWORDS |
