Abstract
With the raise of new biological technologies, as for example DNA chips, and IT technologies (e.g. storage capacities), health care domain has evolved through the last years. Indeed, new high technologies allow for the analysis of thousands of genomic parameters related to various deseases (as cancer, Alzheimer), and how to link them to clinical parameters. In parallel, storage evolutions enable nowadays researchers to gather a huge amount of data generated by biological experiments. This Ph.D thesis is strongly related to medical data mining. We tackle the problem of extracting gradual patterns of the form ''the older a patient, the less his memories are accurate''. To handle different types of information, we propose to extract gradualness for an extensive range of patterns: gradual itemsets, gradual multidimensionnal itemsets, gradual sequencial patterns. Every contribution is experimented on a synthetic or real datasets.