การลดหลักเกณฑ์เชื่อมโยงจากการผ่านตัววัดค่าสนับสนุนแบบอ่อนที่ขึ้นอยู่กับความเชื่อมั่นประพจน์แย้งสลับที่และหลักเกณฑ์กำหนดทิศทาง / ธนาพร ธัญญะเศรษฐ์ = Weak support association rule reduction based on contrapositive confindence and direction setting rules / Tanaporn Tunyaset
Association analysis is one of the data mining techniques that extracts interesting rules or interesting relationship within data. In order to determine interestingness, two measures, the support and confidence are used. Normally, experts must set the appropriate minimum support and minimum confidence to filter uninteresting association rules out. Recently, a pair of measures are presented called the weak support and confidence. The weak support is the measure that describes the probability of events which do not contradict the rule. Its minimum weak support must be set higher than 0.5 since rules that admits more than 50% chance of contradictory events are useless. With the setting of minimum weak support and the minimum confidence, the WS algorithm can extract significant confidence rules. This cause a large number of rules to be generated. To reduce the number of these rules, this research proposes two filtering methods, the minimum contrapositive confidence filter and direction setting rule filter. The CCWS algorithm applies the minimum weak support and the minimum confidence together with the minimum contrapositive confidence while the DSCCWS algorithm additionally applies the direction setting rule concept to the CCWS algorithm. In order to compare among these algorithms, the sensitivity evaluation is selected. Our result shows that DSCCWS algorithm can effectively reduce a significant number of rules while maintains similar sensitivity to the rest of other algorithms.