Anomaly prediction by AI and visualization of prediction rationale

Abstract

In this paper, a new anomaly prediction methodology using AI is described. The prediction rationale is shown using SpecTRM, a formal method.The AI model was built using Random Forest. AI models are usually built as blackbox ones. So, we cannot know the rationale of the inference. However, by using the proposed methodology, the reason of the prediction can be shown by characteristics of the combinations of input data.