The discovery of frequent patterns is one of the problems in data mining. To better understand the influence of the data on the algorithms, we present an experimental study of data sets commonly used by the community. This study lead to a new classification of data based on edge: stable and consistent with the performance of algorithms. Despite the large number of studies and a theoretical framework for extracting interesting patterns problems, the use of these algorithms for solving problems "equivalent" is uncommon and remains difficult. Given these limitations, we propose a generic algorithm for discovering interesting patterns borders, called ABS (Adaptive Search borders), dynamically adapting its strategy to data. In addition, a generic component library C + + has been proposed to facilitate the development of software solutions for this family of problems
