Bayesian structure learning and visualization for technology analysis

Abstract

To perform technology analysis, we usually search patent documents related to target technology. In technology analysis using statistics and machine learning algorithms, we have to transform the patent documents into structured data that is a matrix of patents and keywords. In general, this matrix is very sparse because its most elements are zero values. The data is not satisfied with data normality assumption. However, most statistical methods require the assumption for data analysis. To overcome this problem, we propose a patent analysis method using Bayesian structure learning and visualization. In addition, we apply the proposed method to technology analysis of extended reality (XR). XR technology is integrated technology of virtual and real worlds that includes all of virtual, augmented and mixed realities. This technology is affecting most of our society such as education, healthcare, manufacture, disaster prevention, etc. Therefore, we need to have correct understanding of this technology. Lastly, we carry out XR technology analysis using Bayesian structure learning and visualization.