Title:	FULL MODEL SELECTION IN HUGE DATASETS THROUGH A META-LEARNING APPROACH
Author(s):	Angel Díaz Pacheco and Carlos Alberto Reyes-García
ISBN:	978-989-8533-80-7
Editors:	Ajith P. Abraham, Jörg Roth and Guo Chao Peng
Year:	2018
Edition:	Single
Keywords:	Big Data, Meta-Learning, Model Selection
Type:	Full Paper
First Page:	19
Last Page:	26
Language:	English
Cover:
Full Contents:	click to dowload
Paper Abstract:	One of the main concerns in the pattern recognition field is enhancing the predictive accuracy of the classification algorithms. While some efforts are focused in the development of new algorithms, another approach is centered in the search of the combination of data preparation techniques, a subset of features and, the selection of the adequate learning algorithm with its hyper parameters tuned. This approach is known as Full Model Selection (FMS) and because of the high search space, this paradigm is largely unexplored especially in large size datasets. In this work, the use of the meta-learning approach is suggested as a way to solve the FMS problem in huge datasets and retain the knowledge gained of each analysis performed which would be lost in another way.
	Go Back