XMAP: eXplainable mapping analytical process

Abstract As the number of artificial intelligence (AI) applications increases rapidly and more people will be affected by AI’s decisions, there are real needs for novel AI systems that can deliver both accuracy explanations. To address these needs, this paper proposes a new approach called eXplainable Mapping Analytical Process (XMAP). Different from existing works in explainable AI, XMAP is highly modularised interpretability each step easily obtained visualised. A core algorithms developed to capture distributions topological structures data, define contexts emerged build effective representations classification tasks. The experiments show provide useful interpretable insights across analytical steps. For binary task, its predictive performance very competitive as compared advanced machine learning literature. In some large datasets, even outperform black-box without losing interpretability.