Distant Event Prediction Based on Sequential Rules
| Autor: | Fahed, Lina, Lenca, Philippe, Haralambous, Yannis, Lefort, Riwal |
|---|---|
| Přispěvatelé: | Fahed, Lina, Laboratoire ISEN (L@BISEN), Institut supérieur de l'électronique et du numérique (ISEN)-YNCREA OUEST (YO), Lab-STICC_IMTA_CID_DECIDE, Laboratoire des sciences et techniques de l'information, de la communication et de la connaissance (Lab-STICC), Institut Mines-Télécom [Paris] (IMT)-IMT Atlantique Bretagne-Pays de la Loire (IMT Atlantique), Institut Mines-Télécom [Paris] (IMT)-École Nationale d'Ingénieurs de Brest (ENIB)-École Nationale Supérieure de Techniques Avancées Bretagne (ENSTA Bretagne)-Université de Bretagne Sud (UBS)-Université de Brest (UBO)-Centre National de la Recherche Scientifique (CNRS)-Université Bretagne Loire (UBL)-Institut Mines-Télécom [Paris] (IMT)-IMT Atlantique Bretagne-Pays de la Loire (IMT Atlantique), Institut Mines-Télécom [Paris] (IMT)-École Nationale d'Ingénieurs de Brest (ENIB)-École Nationale Supérieure de Techniques Avancées Bretagne (ENSTA Bretagne)-Université de Bretagne Sud (UBS)-Université de Brest (UBO)-Centre National de la Recherche Scientifique (CNRS)-Université Bretagne Loire (UBL), Département Logique des Usages, Sciences sociales et Sciences de l'Information (IMT Atlantique - LUSSI), IMT Atlantique Bretagne-Pays de la Loire (IMT Atlantique), Institut Mines-Télécom [Paris] (IMT)-Institut Mines-Télécom [Paris] (IMT), Département Informatique (IMT Atlantique - INFO), Arkéa |
| Jazyk: | angličtina |
| Rok vydání: | 2020 |
| Předmět: |
[INFO.INFO-AI] Computer Science [cs]/Artificial Intelligence [cs.AI]
anti-monotonicity [INFO.INFO-LG]Computer Science [cs]/Machine Learning [cs.LG] association rule mining sequence mining [INFO]Computer Science [cs] [INFO.INFO-LG] Computer Science [cs]/Machine Learning [cs.LG] [INFO] Computer Science [cs] Data mining [INFO.INFO-AI]Computer Science [cs]/Artificial Intelligence [cs.AI] distant prediction |
| Zdroj: | Data Science and Pattern Recognition Data Science and Pattern Recognition, 2020, 4 (1), pp.1-23 |
| ISSN: | 2520-4165 |
| Popis: | International audience; Event prediction in sequence databases is an important and challenging data mining task. We focus on the specific case of prediction of distant events. Our aim is to mine sequential association rules with consequents that are temporally distant from their antecedents. We therefore propose two new algorithms: D-SR-postMining and D-SR-in-Mining (D-SR stands for Distant Sequential Rules). The originality of these algorithms is that they integrate a minimal gap constraint between the antecedent and the consequent of existing rules, which, as we prove, has an anti-monotonicity property. This approach allows to predict events with enough time in advance (at least as much as the gap). Both algorithms are designed to coexist with legacy rule mining algorithms: D-SR-postMining can be used as a post-processing step of traditional mining algorithms, and D-SR-inMin-ing can be integrated into the mining process of such algorithms. Experiments on three data sets show that both algorithms are efficient for mining distant rules and scalable on large data sets. Even better, D-SR-inMining reduces execution time significantly (up to 9 times). Furthermore, an in-depth analysis of the rules mined from a real-world bank data set, demonstrates the efficiency of such rules for real-world applications such as churn analysis. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|