Ahmed Halioui, Tomas Martin, Petko Valtchev, Abdoulaye Baniré Diallo
{"title":"基于本体的工作流模式挖掘:在生物信息学专业知识获取中的应用","authors":"Ahmed Halioui, Tomas Martin, Petko Valtchev, Abdoulaye Baniré Diallo","doi":"10.1145/3019612.3019866","DOIUrl":null,"url":null,"abstract":"Workflow platforms enable the construction of solutions to complex problems as step-wise processes made of components including methods, tools, data formats, parameters, etc. Successful workflow solutions require a mastering of the different components paving the way to automated acquisition of problem solving expertise. Thus, process mining could be applied to discover workflow patterns. Due to the combinatorics of component instances in rich domains such as bioinformatics, generalized patterns could be a relevant way of abstraction. Here, we propose an approach for mining workflow patterns, defined on the top of a domain ontology which categorizes workflow elements and their interactions. While original workflows are doubly-labelled DAGs, the underlying problem is transformed into a mining of generalized sequential patterns with links between their items. The proposed mining method traverses the ensuing pattern space using five refinement primitives that exploit the is-a links from the ontology. To assess the prediction power of the approach, we applied the generated patterns as templates in a recommendation platform to complete partial workflows under construction. The analyses of recommendations vs. actual content of a real-world dataset reveals that non trivial patterns can be found and further used to provide plausible recommendations with high accuracies (fMeasure >75+).","PeriodicalId":20728,"journal":{"name":"Proceedings of the Symposium on Applied Computing","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2017-04-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"5","resultStr":"{\"title\":\"Ontology-based workflow pattern mining: application to bioinformatics expertise acquisition\",\"authors\":\"Ahmed Halioui, Tomas Martin, Petko Valtchev, Abdoulaye Baniré Diallo\",\"doi\":\"10.1145/3019612.3019866\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Workflow platforms enable the construction of solutions to complex problems as step-wise processes made of components including methods, tools, data formats, parameters, etc. Successful workflow solutions require a mastering of the different components paving the way to automated acquisition of problem solving expertise. Thus, process mining could be applied to discover workflow patterns. Due to the combinatorics of component instances in rich domains such as bioinformatics, generalized patterns could be a relevant way of abstraction. Here, we propose an approach for mining workflow patterns, defined on the top of a domain ontology which categorizes workflow elements and their interactions. While original workflows are doubly-labelled DAGs, the underlying problem is transformed into a mining of generalized sequential patterns with links between their items. The proposed mining method traverses the ensuing pattern space using five refinement primitives that exploit the is-a links from the ontology. To assess the prediction power of the approach, we applied the generated patterns as templates in a recommendation platform to complete partial workflows under construction. The analyses of recommendations vs. actual content of a real-world dataset reveals that non trivial patterns can be found and further used to provide plausible recommendations with high accuracies (fMeasure >75+).\",\"PeriodicalId\":20728,\"journal\":{\"name\":\"Proceedings of the Symposium on Applied Computing\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2017-04-03\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"5\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Proceedings of the Symposium on Applied Computing\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1145/3019612.3019866\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Proceedings of the Symposium on Applied Computing","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1145/3019612.3019866","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Ontology-based workflow pattern mining: application to bioinformatics expertise acquisition
Workflow platforms enable the construction of solutions to complex problems as step-wise processes made of components including methods, tools, data formats, parameters, etc. Successful workflow solutions require a mastering of the different components paving the way to automated acquisition of problem solving expertise. Thus, process mining could be applied to discover workflow patterns. Due to the combinatorics of component instances in rich domains such as bioinformatics, generalized patterns could be a relevant way of abstraction. Here, we propose an approach for mining workflow patterns, defined on the top of a domain ontology which categorizes workflow elements and their interactions. While original workflows are doubly-labelled DAGs, the underlying problem is transformed into a mining of generalized sequential patterns with links between their items. The proposed mining method traverses the ensuing pattern space using five refinement primitives that exploit the is-a links from the ontology. To assess the prediction power of the approach, we applied the generated patterns as templates in a recommendation platform to complete partial workflows under construction. The analyses of recommendations vs. actual content of a real-world dataset reveals that non trivial patterns can be found and further used to provide plausible recommendations with high accuracies (fMeasure >75+).