Improving ontological knowledge with reinforcement methods in recommendation of the best data mining method for a real environmental problem
Gespeichert in:
| Titel: | Improving ontological knowledge with reinforcement methods in recommendation of the best data mining method for a real environmental problem |
|---|---|
| Autoren: | Gibert, Karina, Sànchez-Marrè, Miquel |
| Weitere Verfasser: | Universitat Politècnica de Catalunya. Departament d'Estadística i Investigació Operativa, Universitat Politècnica de Catalunya. Departament de Ciències de la Computació, Universitat Politècnica de Catalunya. KEMLG - Grup d'Enginyeria del Coneixement i Aprenentatge Automàtic |
| Quelle: | Recercat. Dipósit de la Recerca de Catalunya instname UPCommons. Portal del coneixement obert de la UPC Universitat Politècnica de Catalunya (UPC) |
| Verlagsinformationen: | Asociación Española para la Inteligencia Artificial (AEPIA), 2015. |
| Publikationsjahr: | 2015 |
| Schlagwörter: | reinforcement, Artificial intelligence, knowledge - based recommender system, Programming (Mathematics), ontological knowledge, Matemàtiques i estadística::Investigació operativa::Programació matemàtica [Àrees temàtiques de la UPC], Matemàtiques i estadística::Anàlisi numèrica::Modelització matemàtica [Àrees temàtiques de la UPC], Intel·ligència artificial, data mining, 90C Mathematical programming, 68 Computer science::68T Artificial intelligence [Classificació AMS], Classificació AMS::68 Computer science::68T Artificial intelligence, Programació (Matemàtica), Àrees temàtiques de la UPC::Matemàtiques i estadística::Anàlisi numèrica::Modelització matemàtica, Àrees temàtiques de la UPC::Matemàtiques i estadística::Investigació operativa::Programació matemàtica, recommender system for data mining methods |
| Beschreibung: | There are many data mining techniques available for a user wishing to discover some model from her/his data. This diversi ty can cause some trou- bles to the final non - expert users, who often do not have a clear idea of what are the available methods, and frequently have doubts about the most suitable method for a concrete problem in a domain. In previous works, prior ontologi- c al knowledge about data mining methods has been used to describe the main characteristics of a collection of methods and to filter which methods are suita- ble or not for a given real data mining problem, by matching their characteris- tics with those hold in the target dataset. In this paper, the concept of rein- forcement tables is introduced to move to a multi - criteria scenario in which a measure of relevance is computed for every method . A contribution of the work is to develop an open - frame where both the characteristics of methods consid- ered in the reference ontology and the reinforcement tables may evolve along time according to changes in the methodological state of the art, going beyond classical expert systems. The paper introduces the formal framework and some examples to illustrate the performance of the proposal. |
| Publikationsart: | Conference object |
| Dateibeschreibung: | application/pdf |
| Zugangs-URL: | http://hdl.handle.net/2117/100455 https://hdl.handle.net/2117/100455 |
| Rights: | CC BY NC ND |
| Dokumentencode: | edsair.dedup.wf.002..b2b7f64bf19fb1f777555e0ec522d0f8 |
| Datenbank: | OpenAIRE |
Nájsť tento článok vo Web of Science | Abstract: | There are many data mining techniques available for a user wishing to discover some model from her/his data. This diversi ty can cause some trou- bles to the final non - expert users, who often do not have a clear idea of what are the available methods, and frequently have doubts about the most suitable method for a concrete problem in a domain. In previous works, prior ontologi- c al knowledge about data mining methods has been used to describe the main characteristics of a collection of methods and to filter which methods are suita- ble or not for a given real data mining problem, by matching their characteris- tics with those hold in the target dataset. In this paper, the concept of rein- forcement tables is introduced to move to a multi - criteria scenario in which a measure of relevance is computed for every method . A contribution of the work is to develop an open - frame where both the characteristics of methods consid- ered in the reference ontology and the reinforcement tables may evolve along time according to changes in the methodological state of the art, going beyond classical expert systems. The paper introduces the formal framework and some examples to illustrate the performance of the proposal. |
|---|