EADR: an ensemble learning method for detecting adverse drug reactions from twitter

Adverse Drug Reactions (ADRs) pose a significant public health concern. In recent years, the use of social media data, particularly Twitter, has emerged as a valuable resource for identifying ADRs. The objective is to leverage machine learning and data mining algorithms to extract pertinent information from this platform, with the aim of identifying ADRs that could avert fatalities and hospitalizations. A novel ensemble ADR approach (EADR) is proposed in this study to detect ADRs from Twitter data. The EADR method encompasses several steps, including Twitter data preprocessing, addressing data imbalance through a combination of oversampling and under sampling methods, feature extraction, and the utilization of a Stacking Model, a classification system based on ensemble learning. Experimental results from the Twitter data set demonstrate that the proposed stacking method outperforms its single models in the first level, yielding an 87% F-score, 86% recall, and 87% precision, thus show casing its efficacy in ADR detection.