Meta-Combiner
A Meta-Combiner is an Ensemble Learning Task used with missing attribute values.
- Example(s):
- Counter-Example(s):
See: Learner, Classifier, Data Mining Algorithm, Covering Machine Learning Algorithm.
References
2018
- (Bruha, 2018) ⇒ Ivan Bruha. "S-fold Meta-Combiner for Missing Values Processing: Case Study" (PDF) Retrieved:2018-12-23
- ABSTRACT: Efficient data mining (DM) algorithms have to contain high-performance procedures for processing real-world databases. One of the problems these efficient DM algorithms are faced by are unknown (missing) attribute values in databases. Therefore, robust DM algorithms should comprise some routines for processing these unknown values when acquiring knowledge from real-world databases.
There exist several such routines for each DM paradigm. Quite a few experiments have revealed that each dataset has more or less its own 'favourite' routine for processing unknown attribute values. One possibility how to process efficiently unknown values is exposed in this paper. We use the covering machine learning algorithm CN4 which contains six routines for unknown attribute values processing. Our system runs these routines independently, and afterwards, a meta-learner (meta-combiner) is used to derive a meta-classifier that makes up the overall (final) decision about the class of input unseen objects.
The meta-combiner encompasses in its internal control structure several parameters that are to be set up by the designer or user of the system. One of the crucial parameters is the number S of subsets which a training set is partitioned into during the meta-learning. We are then talking about S-fold meta-combiner. Usually, the ‘foldness’ S is equal to 2 or the size of the training set. This paper exhibits the performance of the meta-combiner for processing unknown attribute values as a function of S. The results of experiments for various values S and various percentages of unknown attribute values on real-world data are presented and analyzed.
- ABSTRACT: Efficient data mining (DM) algorithms have to contain high-performance procedures for processing real-world databases. One of the problems these efficient DM algorithms are faced by are unknown (missing) attribute values in databases. Therefore, robust DM algorithms should comprise some routines for processing these unknown values when acquiring knowledge from real-world databases.
2017
- (Sammut & Webb, 2017) ⇒ Claude Sammut, and Geoffrey I. Webb. (2017). "Meta-Combiner.” In: (Sammut & Webb, 2017). DOI:10.1007/978-1-4899-7687-1
- QUOTE: A meta-combiner is a form of ensemble learning technique used with missing attribute values. Its common topology involves base learners and classifiers at the first level, and meta-learner and meta-classifier at the second level. The meta-classifier combines the decisions of all the base classifiers.
2010
- (Bruha, 2010) ⇒ Ivan Bruha. (2010)."Classification of Datasets with Missing Values: Two Level Approach". In: Proceedings of the 1the 10th International Workshop on Pattern Recognition in Information Systems (ICEIS 2010). ISBN:978-989-8425-14-0 doi:10.5220/0003017800900098
- QUOTE: The algorithm CN4 processes a given database for each of six routines for missing attribute values independently. We can thus view the CN4 algorithm with various routines as independent base learners. Consequently, we obtain (at the base level) six independent base classifiers. Also, a meta-database is derived from the results of the base classifiers, and then a meta-learner induces a meta-classifier. We call the entire system meta-combiner (namely Meta-CN4 to emphasize the origin of the algorithm).
2004
- (Bruha, 2004) ⇒ Ivan Bruha. (2004). "Meta-Learner for Unknown Attribute Values Processing: Dealing with Inconsistency of Meta-Databases". In: Journal of Intelligent Information Systems, 22 (1). ISBN:1573-7675, 0925-9902 doi:10.1023/A:1025880714026----