Predictive Analysis of Pharmaceutical Compounds Using Kernel Naive Bayes in Clinical Informatics

The drug classification into various needful types, improved quality clinical decisions, and more accurate support of pharmacovigilance are some areas of pharmaceutical sciences that can be transformed using Machine learning (ML). Encyclopedia of Information Systems 3rd Edition Kernel Naive Bayes for Drug Classification. This study describes a Kernel Naive Bayes (KNB) model for drug classification based on a wide variety of pharmacological and therapeutic properties. From drug product data repository, this model integrates at least fundamental drug-related features, such as dosage forms, routes of administration, adverse reactions, interactions, and indications for use, which are considered as basic elements in pharmaceutical research and clinical pharmacy. It uses a Gaussian kernel to model continuous variables and a Multivariate Multinomial (MVMN) distribution to model categorical features — which allows for a more complex relationship among the features. To improve interpretability and mitigate noise, irrelevant or sparse attributes (i.e., regulatory codes, precautionary labels) were excluded. The last model attained an accuracy of 83.2% along with a prediction speed of ~1600 observations/sec proving its potential in handling largescale pharmaceutical data effectively and efficiently. These results support the relevance of kernel-based probabilistic models in pharmacy-related issues, especially in drug safety screening, automated classification, and pharmacological data mining.