Perception Based Associations in Time Series Data Bases

The paper discusses different aspects of the development of perception-based decision making systems. These systems are based on inference procedures transforming associations extracted from the time series data bases into generalized-constraint inference rules. Different types of simple and composite perception based constraints are discussed. Various measures of association between time series in the presence of perception based constraints are considered: association rules, association rules with perception based frequencies, correlation rules, and local trend associations based on moving approximations. Finally, the methods of transformation of these associations into the inference rules that can be used in perception based reasoning are proposed