Doubts about Min Confidence Threshold
对最小置信度门限的置疑

The two thresholds: min support and min confidence are often used to evaluate if a rule is a strong rule or not in association rules mining. The present paper analyzes and explores the practical significance of min confidence both theoretically and practically. It finds that not only the quality of the mined rules is comparatively low but also some important rules are probably missed out if min confidence is used to restrict. Different from previous research,this paper proposes that upgrade rate can reflect the actuality more accurately than min confidence and the result will be more accurate and clearer if min support and min upgrade are taken as the weighing rule in association rules mining.