Information gain-based selection of sequential patterns extracted from partial unimodal probabilistic bases of sequences

Nicolas Méger (1) , Tuan Nguyen (1) , Christophe Rigotti (2, 3, 4) , Catherine Pothier (5) , Emmanuel Trouvé (1)

1 LISTIC - Laboratoire d'Informatique, Systèmes, Traitement de l'Information et de la Connaissance
2 LIRIS - Laboratoire d'InfoRmatique en Image et Systèmes d'information
3 DM2L - Data Mining and Machine Learning
4 BEAGLE - Artificial Evolution and Computational Biology
5 imagine - Extraction de Caractéristiques et Identification

The uncertainty of symbolic data can be represented by probability mass functions. Numerous work adopt this approach to characterize the uncertainty of the events forming a probabilistic base of sequences and extract sequential patterns under the possible worlds semantics. To our knowledge, there is no method for selecting sequential patterns from probabilistic bases of sequences whose probability mass functions are unimodal and for which only the probabilities of the modes are available. Since this situation arises for several kinds of data, a method for selecting sequential patterns extracted from partial unimodal probabilistic bases of sequences is thus proposed in this paper. Using an information gain approach, it outputs informative patterns whose occurrences tend to describe the dataset in a complementary way. Experiments on synthetic and real datasets show that the method is scalable and that selected patterns, beside being informative and complementary, help end-users to complete their knowledge.

Mots clés: Uncertain data; Sequential pattern mining; Probabilistic databases; Information gain

Domaines: Apprentissage [cs.LG]; Intelligence artificielle [cs.AI]

Nicolas Méger, Tuan Nguyen, Christophe Rigotti, Catherine Pothier, Emmanuel Trouvé. Information gain-based selection of sequential patterns extracted from partial unimodal probabilistic bases of sequences. 2024-1, Laboratoire d'Informatique, Systèmes, Traitement de l'Information et de la Connaissance (LISTIC) - Polytech Annecy-Chambéry. 2024, pp.85. ⟨hal-04472843⟩