Development of the Python3 based Bayesian inference dynamic visualization tool “B.T.V.T.I.” with multiplatform support designed to assist decision-making in healthcare and welfare

Abstract

Objective：To address the three pending issues that were raised after the release of B.I.T. (Bayesian inference tool). A Bayesian inference visualization tool expands the user experience by providing support to multiple platforms; it adds robustness to calculations instead of approximation and offers various inference visualization options other than the normal distribution.

Method：Using Python3, NumPy, SciPy, and the plotting tool Matplotlib, we created a new Bayesian inference visualization tool that runs on multiple operating systems and assures a robust calculation of the posterior probability. In addition, new cases of Bayesian inferences other than the normal distribution were collected for visualization.

Results：The newly developed B.T.V.T.I. (Bayesian tool for various types of inferences) runs on Windows, macOS, and Linux (Ubuntu). It is a software program with twelve Bayesian inference visualization screens, out of which nine have the statistical function implementation of SciPy. The program calculates and displays graph curves of the posterior distribution and posterior probability values at the same time as the observed values and variables are input. However, because a different programming language is used, its operation screen is different from that of B.I.T.

Discussion：The future expandability of B.T.V.T.I. to other devices, the application and limits of SciPy, and future implementation and visualization of the uncited Bayesian inference cases were discussed. Ideally, users must be familiar with the principle, premises, and applications of the Bayesian inference to use this software.