Bitwise Vertical Mining of Minimal Rare Patterns

Rare patterns are essential forms of patterns in many real-world applications such as interpretation of biological data, mining of rare association rules between diseases and their causes, detection of anomalies. However, discovering rare patterns can be challenging. In this paper, we present an efficient algorithm for mining minimal rare patterns from sparse and weakly correlated data. The algorithm non-trivially integrates and adapts vertical frequent pattern algorithm VIPER to discover minimal rare patterns in an efficient manner. Evaluation results on our algorithm RP-VIPER show its superiority over existing horizontal rare pattern mining algorithms. Results also highlight the performance improvements brought by our optimized strategies.