Knowledge graph-based rich and confidentiality preserving Explainable Artificial Intelligence (XAI)

Abstract

The paper proposes a novel architecture for explainable artificial intelligence based on semantic technologies and artificial intelligence. We tailor the architecture for the domain of demand forecasting and validate it on a real-world case study. The explanations provided result from knowledge fusion regarding concepts describing features relevant to a particular forecast, related media events, and metadata regarding external datasets of interest. The Knowledge Graph enhances the quality of explanations by informing concepts at a higher abstraction level rather than specific features. By doing so, explanations avoid exposing sensitive details regarding the demand forecasting models, thus preserving confidentiality. In addition, the Knowledge Graph enables linking domain knowledge, forecasted values, and forecast explanations while also providing insights into actionable aspects on which users can take action. The ontology and dataset we developed for this use case are publicly available for further research.