Authors:
Silviu Cacoveanu
1
;
Camelia Lemnaru
1
and
Rodica Potolea
2
Affiliations:
1
Technical University of Cluj-Napoca, Romania
;
2
Techincal University of Cluj-Napoca, Romania
Keyword(s):
Meta-learning Framework, Data Set Features, Performance Metrics, Prediction Strategies.
Related
Ontology
Subjects/Areas/Topics:
Artificial Intelligence
;
Artificial Intelligence and Decision Support Systems
;
Biomedical Engineering
;
Business Analytics
;
Data Engineering
;
Data Mining
;
Databases and Information Systems Integration
;
Datamining
;
Enterprise Information Systems
;
Health Information Systems
;
Informatics in Control, Automation and Robotics
;
Intelligent Control Systems and Optimization
;
Knowledge Engineering
;
Knowledge Engineering and Ontology Development
;
Knowledge-Based Systems
;
Knowledge-Based Systems Applications
;
Sensor Networks
;
Signal Processing
;
Soft Computing
;
Software Engineering
;
Symbolic Systems
Abstract:
Finding the best learning strategy for a new domain/problem can prove to be an expensive and time-consuming process even for the experienced analysts. This paper presents several enhancements to a meta-learning framework we have previously designed and implemented. Its main goal is to automatically identify the most reliable learning schemes for a particular problem, based on the knowledge acquired about existing data sets, while minimizing the work done by the user but still offering flexibility. The main enhancements proposed here refer to the addition of several classifier performance metrics, including two original metrics, for widening the evaluation criteria, the addition of several new benchmark data sets for improving the outcome of the neighbor estimation step, and the integration of complex prediction strategies. Systematic evaluations have been performed to validate the new context of the framework. The analysis of the results revealed new research perspectives in the met
a-learning area.
(More)