Abstract
Recently, with the constant progress in software and hardware technologies, real-world databases tend to grow larger and the hypothesis that a database can be efficiently analyzed by a human being is no longer valid. In order to cope with this new challenge, a new set of methods and technologies (i.e., Knowledge Discovery and Delivery Data Mining or KDD) were developped to enable data exploration, data analysis and data visualisation of very large databases at a high level of abstraction. Furthermore, the data stream model, a new model of data as a continuous and potentially infinite flow, call for new approaches for the pre-processing and data mining operations.<br />In this thesis, we focus on a very specific data mining method: sequential pattern mining. This method aims at discovering correlations among events through time. We present very important theoretical results on the possible concise representations for sequences. Furthermore, we introduce a new pattern called conjunctive sequential pattern. Sampling methods and approaches for sequential pattern mining over data streams and static databases are also introduced and discussed. Moreover, we define an algorithm that allows the discovery of multidimensional sequential patterns. These approaches are also validated with experiments on real data.