Automatic Recommendation of Classifier Ensemble structures using Meta-learning
Machine Learning, Classifier ensembles and Meta-learning
Today we are constantly concerned with classifying things, people, and making decisions, which when we encounter problems with a high degree of complexity, we tend to seek opinions from others, usually from people who have some knowledge or even, as far as possible. possible, be experts in the field of the problem in question, so as to effectively assist us in our decision-making process. In analogy to classification structures, we have a committee of people and or specialists (classifiers) that makes decisions, and based on these answers, a final decision is made (aggregator). Thus, we can say that a committee of classifiers is formed by a set of classifiers (specialists), organized in parallel, that receive input information (pattern or instance), and make an individual decision. Based on these decisions, the aggregator chooses the final single decision of the committee. An important issue in designing classifier committees is the definition of their structure, more specifically, the number and type of classifiers, and the method of aggregation for the highest possible performance. Generally, an exhaustive testing and evaluation process is required to define this structure, and trying to assist with this line of research, this paper proposes two new approaches to automatic recommendation systems of the classifier committee structure, using meta-learning to recommend three of these parameters: the classifier, the number of classifiers, and the aggregator.