Discovering patterns in sequences of events

摘要: Abstract Given a sequence of events (or objects), each characterized by set attributes, the problem considered is to discover rule characterizing and able predict plausible continuation. The rule, called sequence-generating , nondeterministic in sense that it does not necessarily tell exactly which event must appear next sequence, but rather, defines events. basic assumption methodology presented here depends solely on attributes previous sequence. These are either initially given or can be derived from initial ones through chain inferences. Three models employed guide search for rule: decomposition, periodic, disjunctive normal form (DNF). process involves simultaneously transforming sequences instantiating find best match between instantiated model A program, SPARC/E, described implements most as applied discovering generating rules card game Eleusis. This game, scientific discovery, used source examples illustrating performance SPARC/E.