PIS: A probabilistic inference system

A method is proposed for probabilistic inference through empirical observations involving categorical data. This method can detect statistically independent patterns inherent in a set of observed events. The evidence provided by the patterns for or against some hypotheses generated during the inference process are then quantitatively estimated and combined to find the most plausible hypotheses. The proposed method has been implemented for the Probabilistic Inference System (PIS). Because it can detect patterns in observed or inferred events which may not be directly observable, the PIS can be used to aid decision-making in the presence of uncertainty. It has been tested with simulated as well as real-life data, and the results are very satisfactory.