Machine Learning for Drug Recommendation Based on Opinion Analysis of Drug Reviews
摘要
Access to reliable medical information in the digital era remains a challenge due to the dispersion of data, making it difficult to improve the health of users. Medication errors are a major concern, highlighting the need for integrated health recommendation systems. This study aims to use advanced machine learning techniques for the development of a drug recommendation system that takes into account parameters such as age and gender for a personalized and efficient recommendation. Two models Random Forest (RF) and Naive Bayes (NB) were used to classify the comments. The performance of each model was evaluated by analyzing four metrics (accuracy, F1 score, Recall, and precision). The models achieved an accuracy of 0.66 (NB) and 0.77 (RF) when evaluated with three categories of comments (positive, negative, and neutral). However, the accuracy of the model increased when neutral comments were removed, obtaining a value of 0.76 for Naive Bayes and 0.84 for Random Forest. This makes Random Forest (RF) the best model for comment classification.