Single and ensemble-based classifiers in biometric recognition

In the area of machine learning the concept of combining classifiers is proposed as a new direction for the improvement of the performance of individual classifiers. In machine learning, classification is the problem of identifying to which of a set of categories a new observation belongs, on the basis of a training set of data containing observations whose category membership is known. Classification and clustering are examples of pattern recognition, which is the assignment of some sort of output value to a given input value. An algorithm that implements classification in a concrete implementation is known as a classifier. The goal of classification result integration algorithms is to generate more certain, precise and accurate systems. This strategy brings a good solutions in biometrics as well as in medicine, where classifiers support the diagnosis process. Classifiers operate on the set of features and these features have to be reliably selected. Additional, uncorrelated errors made by individual classifiers can be overcome by voting. During lecture advantages as well as disadvantages of single and ensemble-based classifiers will be shown. For example, ensemble of classifiers outperforms any single classifier but learning time of a single classifier is mostly shorter. During the speech, representative examples of biometric cases as well as medical diagnosis will be also given and it will be compared with various classification methods. It will be also noted that features which are used in classifiers training stage can be unbalanced and even can be incomplete. Missing data can create problems for analyzing data, imputation is seen as a way to avoid pitfalls involved with listwise deletion of cases that have missing values. Influence of some data imputation methods on classifiers accuracy will be also performed.