Association rules: Normalizing the lift

Association rules is a popular data mining technique for discovering relations between variables in large amounts of data. Support, confidence and lift are three of the most common measures for evaluating the usefulness of these rules. A concern with the lift measure is that it can only compare items within a transaction set. The main contribution of this paper is to develop a formula for normalizing the lift, as this will allow valid comparisons between distinct transaction sets. Traffic accident data was used to validate the revised formula for lift and the result of this analysis was very strong.