ABSTRACT
Patients administered with more than one drug need to be monitored for increased or decreased effects due to the given dosage. DDIs (drug-drug interactions) are significant operations for patient's safety and efficient management in healthcare. DDI characterisation is extremely important to avoid undesirable drug reactions. Although many techniques have been proposed for understanding DDIs, they fail to provide sufficiently detailed information on DDIs that can help reduce its effects on patients. Hence, this work proposes a framework called ICM (improved classification model) for accurate predictions of DDI types based on input drug details. This work preprocesses data by eliminating missing features using min-max method, thus producing better features that are subsequently chosen using ABC (artificial bee colony) algorithm. The selected features are then used for predicting DDIs using ICM. This work also uses DBNs (deep belief networks) for classifying data. Further, IBA (improved bat algorithm) is used in optimisation of this work. The proposed DBN-IBA model is simulated on MATLAB that shows the model's effectiveness by displaying better DDI prediction rates. This work also recommends guidelines to be followed while developing drugs.
