Sentimental Analysis on Health-Related Information with Improving Model Performance using Machine Learning

Social media platforms are extensively used in exchanging and sharing information user experience, thereby resulting massive outspread viewing of personal experiences many fields life. Thus, informative health-related videos on YouTube highly perceptible. Many users tend to procure medical treatments from social particularly when searching for chronic illness treatments. Sometimes, these sources contain misinformation that cause fatal effects the users’ health. sentimental analyses classifications have been conducted study post comments life science fields. However, no has analysis Arabic comments, which provide details herbal people with diabetes. Therefore, this proposes a model detect discover emotions/opinions treatment is proposed through an by using machine learning classifiers. In addition, new Dataset Herbal Treatments Diabetes (ADHTD), based several videos, introduced. This examines impact four representation methods ADHTD show performance These remove repeating characters dialect character extension known as ‘TATAWEEL’ or ‘MAD’, stemming words, stop words removal N-grams words. Experiments aforementioned handle imbalanced dataset identify best classifiers over textual data. The achieved higher accuracy reached 95% Synthetic Minority Oversampling TEchnique (SMTOE) techniques balanced than dataset.