DIAGNOSIS WITH CONTINUOUS AND DISCRETE CAUSAL RELATIONSHIPS: KNOWLEDGE REPRESENTATION

Abstract

Most present research on diagnostic reasoning has dealt with discrete causal relationships. However, continuous causal relationships are frequently encountered in many applications, particularly in equipment/instrument diagnostic applications. This paper offers an integrated model of discrete and continuous causal relationships based on probability theory. Four types of continuous and discrete causal relationships are identified. For each type of causal relationship the following issues are addressed: its semantics and causation strength based on probability theory, and how causation strength can be acquired. It is also demonstrated that causation probabilities for continuous causal relationships can be derived based on classical statistical theory and Bayesian probability theory. It will be shown in a companion paper that the knowledge representation framework presented in this paper can be used in causal network-based diagnostic inference models.