Automated Adverse Drug Reaction detection in social media datasets, using deep learning methods (HAN, FastText and CNN)

Abstract

Objective: Apparently, the health-related studies have been recently in media spotlight. Social media, like Twitter, has become the valuable online tool to describe early detection of various adverse drug reactions (ADR). Different medications have adverse effect on various cells and tissues, some times more than one cell components would be adversely affected. These types of side effect are occasionally associated with direct or indirect influence of prescribed drug dosage but do not have generally unfavorable genetic based consequences on patients. Materials and Methods: This study aimed to determine a quick and accurate method to collect and classify information related to the distribution of evidenced data on Twitter. We selected “ask a patient” dataset and combination of Twitter “Ask a Patient” datasets including: 6,623, 26,934 and 11,623 reviews. We used deep learning methods with the word2vec to classify ADR comments posted by users and present an architecture by HAN, FastText and CNN. Results: NLP deep learning is able to address more advanced peculiarity in learning information compared to other type of machine learning. Additionally, the current study highlighted the advantage of incorporating various semantic features including topics and concepts. Conclusion: Our approach predicts drugs safety with a high accuracy of 93% (combination of Twitter and "Ask a Patient" datasets) in binary manner. Despite the obvious benefit of various conventional classifiers, deep learning-based text classification methods are shown to be accurate and effective tools for ADR detection.