Classifier Fusion Based on Weighted Voting - Analytical and Experimental Results
Michał Woźniak · 2008
The multiple classifier systems are nowadays one of the most promising directions in pattern recognition. There are many methods of decision making by the group of classifiers. The most popular are methods that have their origin in vote methods, where the decision of the common classifier is a combination of simple classifiers decisions. There exists a trend of combined classifiers, which are making their decisions basing on the discrimination function, this function is a combination of above-mentioned simple classifier functions. This work presents an attempt to estimate the classifier error, which bases on the combined discrimination function. Obtained from this estimation conclusions will serve to formulate project guidelines for this type of decision-making systems. At the end experimental results of combining algorithms are presented, both from computer generated data and for real problem from the medical diagnostics field.