Smart Drug Recommendation System Using Sentiment Analysis of Drug Reviews

Abstract

Global healthcare disruptions have precipitated extraordinary challenges within medical service delivery systems, culminating in severe healthcare professional shortages and compromised patient access to qualified therapeutic consultation. These circumstances have fostered widespread autonomous medication selection behaviors, frequently resulting in inappropriate or potentially detrimental treatment protocols. Our investigation presents a novel therapeutic advisory framework that harnesses user experience analytics through sophisticated computational intelligence methodologies to furnish personalized and trustworthy pharmaceutical guidance. Our innovative approach transcends conventional computational learning paradigms by establishing a comprehensive digital consultation platform that exhibits therapeutic options alongside exhaustive performance analytics and integrated patient insight compilation. We exploited the extensive Drugs.com user experience database, implementing cutting-edge textual processing methodologies encompassing Term Frequency-Inverse Document Frequency computation, Word2Vec semantic modeling techniques, and sophisticated attribute extraction protocols. Our research examined multiple computational approaches including Logistic Regression modeling, Random Forest classification frameworks, Naive Bayes probabilistic models, and Support Vector Machine architectures for emotional content analysis. The optimized TF-IDF methodology coupled with Linear Support Vector Classification generated outstanding results, attaining 93% classification accuracy. Our user-centric interface exhibits pharmaceutical options through innovative card-based presentations, featuring therapeutic names, performance indicators, comprehensive descriptions, and balanced summaries highlighting therapeutic benefits alongside potential adverse effects. This groundbreaking design amplifies user comprehension while facilitating evidence-based decision-making rooted in authentic patient testimonials, providing crucial support for patients and healthcare providers, especially within geographically isolated regions experiencing healthcare access limitations.

Introduction

Modern healthcare faces critical challenges, including inadequate medical workforce distribution in rural areas and frequent errors in medication prescriptions. With ongoing pharmaceutical innovations, medical professionals struggle to stay updated on effective treatments. Artificial intelligence (AI) offers promising solutions, and this project leverages AI to improve pharmaceutical recommendations by analyzing patient feedback.

The system integrates Google’s Gemini API for advanced natural language processing, alongside machine learning classifiers (such as Support Vector Classification) trained on a large drug review dataset to assess the sentiment of patient feedback. The framework is built as a three-layer architecture:

1. Patient Interaction Module: A web-based interface allows users to input medical conditions and receive drug recommendations displayed as interactive cards, featuring drug names, effectiveness scores, and balanced patient feedback highlighting both benefits and limitations.

2. Backend Processing Infrastructure: This layer processes queries using the Gemini API for language understanding and multiple machine learning models to classify sentiment in patient reviews, achieving up to 93% accuracy.

3. Information Repository: A cleaned and preprocessed Drug Review Dataset from Drugs.com serves as the knowledge base, with sentiment-labeled reviews used to train classifiers.

Evaluation shows that the system delivers highly accurate sentiment classification and offers an intuitive, patient-friendly interface that transparently presents both positive and negative drug information. This balanced approach empowers patients, especially those in underserved areas, to make informed decisions about their medications without relying solely on healthcare professionals. The integration of advanced NLP through Gemini API further enhances recommendation relevance and contextual understanding.

Conclusion

This investigation introduced an intelligent pharmaceutical consultation assistant integrating emotional analysis, computational learning classification algorithms, Google\'s Gemini API, and interactive web interfaces. Our system achieved robust performance metrics (93% accuracy in sentiment prediction) while demonstrating practical utility by providing results in accessible formats. The capability for presenting both positive and negative pharmaceutical perspectives creates balanced and transparent recommendations, enabling patients to make informed decisions without relying entirely on professional consultations. This proves particularly valuable for individuals in remote or underserved regions with limited healthcare access. Future enhancements may incorporate real-time patient health data, additional pharmaceutical databases, and advanced deep learning architectures to further improve recommendation quality and expand therapeutic coverage.

References

[1] J. Smith, L. Zhang, and A. Taylor, \"Comparative Analysis of Legacy VPN Protocols: PPTP, L2TP/IPSec, and SSL,\" Journal of Network Security, vol. 9, no. 2, pp. 101–109, Mar. 2017. [2] D. Patel and S. Rao, \"Enterprise VPN Framework using IPSec: Security and Deployment Challenges,\" International Journal of Information Security Research, vol. 6, no. 4, pp. 223–230, 2018. [3] M. K. Pal, R. Kumar, and V. Sharma, \"Sentiment Analysis of Drug Reviews Using Machine Learning Techniques,\" International Journal of Advanced Computer Science and Applications, vol. 11, no. 5, pp. 456–462, 2020. [4] N. Hussain, M. Afzal, and K. Malik, \"A Survey on Sentiment Analysis Techniques in Healthcare,\" IEEE Access, vol. 9, pp. 138–156, 2021. [5] C. Vens and F. Costa, \"Healthcare Recommender Systems: A Survey,\" Artificial Intelligence in Medicine, vol. 81, pp. 1–28, 2017. [6] UCI Machine Learning Repository, \"Drug Review Dataset (Drugs.com),\" Available: https://archive.ics.uci.edu/ml/datasets/Drug+Review+Dataset+(Drugs.com). [7] Google, \"Gemini API Documentation,\" Google AI, 2024. Available: https://ai.google.dev/gemini-api. [8] T. Mikolov, K. Chen, G. Corrado, and J. Dean, \"Efficient Estimation of Word Representations in Vector Space,\" in Proc. International Conference on Learning Representations (ICLR), 2013. [9] N. Topalidou, E. Andrikopoulou, and G. Antoniou, \"GalenOWL: A Semantic-Enabled Drug Recommendation Framework,\" Journal of Biomedical Informatics, vol. 62, pp. 1–12, 2016. [10] J. Han, M. Kamber, and J. Pei, Data Mining: Concepts and Techniques, 3rd ed. Burlington, MA: Morgan Kaufmann, 2011.

Copyright

Copyright © 2025 Prof. Dr. S D N Hayath Ali, Dr. Girish Kumar D, Mrs. Laksmi S, Prof. Subhashree D Lakshmi S, Prof. M M Harshitha. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.