Medicine recommendation system based on review sentiment analysis and similarity metrics

A recommendation system can help the user combine requirements and make well-informed decisions based on a wealth of complex information. Recommendations derived from sentiment analysis appear to be extremely difficult because user-generated information is expressed in a variety of intricate ways utilizing human language. There has been not a lot of research on health and medical concerns, with most studies concentrating on popular topics like restaurant, movie, and electrical product reviews. Sentiment analysis concerning medical care in general, and individual medication experiences in particular, may provide a great deal of insight into how to prioritize enhancing public health and make the best decisions. This study, create a framework to a medication recommender system that used content-based recommendation techniques and technologies of sentiment analysis for medication reviews. The purpose of this study is to design a medicine recommender model that assists medical healthcare workforces to provide healthcare services securely by giving relevant medicine to patients. Firstly, we apply the sentiment analysis step to each of including removing the stop word, stemming, and lemmatization, and convert each review to a numerical form by applying the Count Vectorizer. Secondly, we implement different similarity measurements on these vectors. After this, we will choose one of these similarity algorithms that gave us the best results and sort the results in descending order, which will lead to a list of suitable medicines. After using different metrics to find the similarity, the model built on cosine similarity, and has the highest average score The result was two out of three health medications, and there are also cases all recommendations are correct.