Prediction of harmful drug reaction is the most focused research concern in the field of medical for predicting drug performance and safety. Millions of user’s expressions about drugs can be found in multiple sources on the Internet. Analyzing these results, drug interactions can be accurately predicted and therefore the necessary measures to improvise the safety factors of drugs can be taken accordingly. Many researchers are developing different research methods to predict the drug reactions. Among them ADR mine is the note-worthy implementation which uses the features of sentimental analysis. Nevertheless, in the existing work, the quality of classification would be compromised if there were a variety of irrelevant words found in the input data. Also the Conditional Random Field (CRF) classifier is used for classification, but it is highly computationally complex at the training stage of the algorithm. This makes it very difficult to retrain a model as newer data is available. These issues are addressed in the proposed system, namely the Semantic Drug Reaction Prediction Framework (SDRPS). Preprocessing is initially performed in the proposed research work to perform filtering of stop words. Features of the sentimental analysis are extracted as provided in existing work. Ultimately prediction of drug effects is rendered using hybridized SVM and ANN classifier which is computationally better than Conditional Random Fields. Lastly, k means clustering is used to group both positive and negative comments. The performance of the research work is assessed by taking drug related comments in the MATLab tool and revealed to perform in an improved manner than the current ADR mine.
Volume 12 | 03-Special Issue
Pages: 891-899
DOI: 10.5373/JARDCS/V12SP3/20201332