Knowledge Graphs for Explainable Big Data Decision Making

Abstract

The rapid growth of big data within sectors has transformed how organizations make strategic, operational and real time decisions. But the big data is unstructured, heterogeneous, and complex and thus incredibly difficult to extract actionable insights. Traditional machine learning models are considered black-boxes despite being very powerful, which restricts any level of trust and interpretation in the decision making process. Knowledge graphs (KGs) have emerged as a promising paradigm to learn, organize and reason over large-scale heterogeneous data in response. KGs are capable of providing context-awareness, semantic understanding and explainability in analytics of large-scale data, by providing an explicit model of the relations between entities. The article describes in detail how knowledge graphs are used to explain big data decision making. It talks about the basis of KGs and how these can be applied to connect structured and unstructured information and how they can be combined with machine learning and AI so as to be understandable. The hospital, financial sector, smart city, and supply chain management among others are just a few examples of how KGs can be applied to promote trust and accountability. Other methodologies evaluated in the article include ontology based modelling, graph embeddings and hybrid KG-deep learning architectures. The findings show that knowledge graphs possess undeniable benefits in terms of transparency and reasoning, yet remain a challenge regarding scale, dynamism and standardization. This paper aims to explain why knowledge graphs are necessary to deliver explainable and reliable AI-based decision systems in the age of big data. The future trend is to develop automated KG building pipelines, communicate with natural language processing (NLP), and collaborate with federated KG models to collaborate with other organizations.