Tucker inference learning method for knowledge graph decision making

Jicang Xu1, Ming Li1, Cheukhang Leung2
1School of Economics and Management, China University of Petroleum, Beijing, 102249, China
2Hong Kong Institute of AI for Science, City University of Hong Kong, Hong Kong

Abstract

This study aims to build a framework called Tucker Reasoning Learning Method to train the upper layer knowledge graph (KG) making explainable and reasonable decisions. The numerical experiments show that the accuracy is 84%. The contributions are as follows: (i) It overcomes in-explainable problems of using deep learning method; (ii) It has more feedback rings and reasonable paths than decision tree method; (iii) Compared with RESCAL’s application in reasoning domain, it enhances 22 percentage points. It is suitable for application scenarios like financial, justice, and medical decision-making, which require explainable and reasoning paths. This study builds a framework called Tucker Reasoning Learning Method to train the upper layer knowledge graph to make explainable and reasonable decisions. The method has the accuracy of 84%, which enhances 22 percentage points compared to the SOTA methods.

Keywords: Knowledge graph, Tensor analysis, Tucker reasoning, Decision making