On the use of the Rademacher complexity in mining sequential patterns

A sequential pattern is a sequence of sets of items. Mining sequential patterns from very large datasets is a fundamental problem in data mining. This thesis formally proves the first rigorous and efficiently computable bound on the Rademacher complexity of sequential patterns. This result is then applied to two key tasks: mining frequent sequential patterns from a given dataset using progressive sampling, and mining true frequent sequential patterns from an unknown generative process.