00040 ¼3020594679a2c410532d6922f3510a

39

A Rule-Based Approach to Multiple Statistical Test Analysis

_Ru]e0I •*•*•••*•••••*•*•••*•*•*•*•*****'

RuleOl — de term mes if incoming signal has lead CUSUM’s additional value less than 0.5. If so, classifies as a false alarm

DOCO	.Written by: .DatÄ™: 4/14/90 ;Compiled: ;Fixes: nonÄ™	WadÄ™ Molnau 4/14/90
NAME	(p ruleOl
	{(goal		^Aname evaluate)
	(task	^Aname	ruleset1)
	(values	^name	additional
		^Acusuml-val {<cl-val>
	(modify <goal-e>
		^Aname	confirm)
THEN	(make task	'name	recommend
		'Nalue	misfire)
	(make task	'Yiame	fired
		''yalue	ruleOl)
CLOSURE	)		ruleOl *************

FigurÄ™ S. Example of OPS5 Production Rule.

procedurÄ™ exists for each rule. If the criteria is met for that rule, the procedurÄ™ fires and the signal is classified as a detection or false alarm. A finaÅ‚ procedurÄ™ exists which passes the data to Ruleset 2 if no classification is madÄ™. The other rulesets operate in the same manner, with the exception of Ruleset 4. AU signals are classified upon leaving the last set of procedures.

The fourth goal is the confirm stage. After receiving a classification for the signal, the system determines if the classification is correct. The confirm procedures can be based on decision rules, or can be user based. For the later, a user's input would be reÄ…uired to continue with the Processing. The user will be notified of the decision and the basis for the decision. After investigation, the user can confirm or contradict the system's fmdings. It is then up to the user to take any reÄ…uired corrective to solve the process problem, if a detection is deemed to be correct.

The finaÅ‚ goal is adjust. Here statistics are kept on performance of the system, rulesets, and individual rules based on the feedback of the confirm stage. Using the feedback and with modifications to the system, this is where any changes in the system parameters and structure could be conducted.