Identification of Association between Prescription Drugs and Side Effects by Analyzing Social Network Messages

Abstract

In this world of internet and social media all people have started discussing about their health information and treatment procedures in the health forums and social media like twitter. Researches are now being focused towards identifying hazardous effects of the prescription drugs and the treatment process through mining this information posted over the internet. Specifically, Twitter can be considered as an important source of information for the detection of such as Adverse Drug Reaction (ADR). The mining or analysis of Twitter messages is not easy because they are of short length, unstructured and almost in informal form. The twitter messages related to drugs prescribed for cardio vascular and diabetes were considered and collected to form the initial dataset. Later they are preprocessed to remove redundancy and improve the further classification process. A set of feature like Semantic, Z-Score, lexicon related features were extracted from the collected tweets to from the training dataset. Next the feature selection is performed using the Pointwise Mutual Information (PMI) approach. Finally, the selected feature set is utilized to train the Support Vector Machine (SVM). The SVM is validated with a test dataset and its performance was found satisfactory when linear function is used as the kernel. This model can be utilized further to identify the association between prescribed drugs and adverse effects from the Tweets and other messages of health forums.