Estimating the serial combination's performance from that of individual base classifiers

Although the large number of MCS topics, serial fusion of multiple classifiers has been poorly investigated so far. In this paper, we propose a model which, starting from the performance of individual classifiers and the traditional hypothesis of decision independence given the class, is able to estimate the performance, in terms of error rates, of the whole serial classification scheme. The model is tested on a large set of data sets and classifiers, and the importance of the basis hypothesis is evaluated under different scenarios, which can be in agreement or not with such hypothesis.