Enhancing Decision-Making in Product Development: Predicting A Medicine-Based Treatment for A New Disease Using A Multidimensional Neural Network

As the COVID-19 pandemic outbreaks all over the world, several clinical medical centers are currently testing numerous medicines in clinical trials. However, there is no powerful medication recommended so far, in addition to the possibility of clinical trials failure in the context of unbeneficial investments and undesired outcomes.

To address this gap, a new quantitative approach has been proposed in the current paper to monitor the medicine-based treatment development. Precisely, the cure and mortality rates of a potential treatment can be predicted systematically by using convenient predictive multidimensional neural networks and based on the quantitative analysis of Big Data for different indicators. More specifically, the diseases’ symptoms, along with the corresponding medicines, the active ingredients that compose these medicines, the related patents and publications data, the related reports in clinical trials, etc. can be all considered as effective key performance indicators in evaluating medicine-based product success. The present methodology is applied on two candidate diseases: “Multiple Myeloma” for the testing phase and “COVID-19” for the prediction phase, using three types of Multidimensional Neural Networks for comparison purposes: “Wide model” (WNN), “Deep model” (DNN) and “Wide and Deep model” (WDNN). The findings show that the WDNN model achieves higher prediction accuracy and outperforms WNN and DNN, with a significant prediction accuracy equal to 98.79%. Consequently, addressing this estimation represents a decision support for pharmaceutical and clinical medical centers and a crucial prerequisite step before proceeding with investments decisions in clinical trials and medicines production.