Advances in Supervised Classification Models

Supervised classification models allow to make predictions about a target random variable, which describes a feature of interest of some object, using the evidence provided by a set of descriptives variables about this object. For example, predict if an email is spam or not using the words it contains. The first part of this book presents two new contributions to the state of the art of these models. Firstly, a new semi-Naive Bayes classifier is presented which provides competitive predictions while requires very low memory resources. These properties make this model attractive for integration into devices with limited memory resources (e.g. mobile phones). The second contribution was a new Bayesian approach to the problem of learning classification trees. In the second part of this book, new supervised classification models based on the selection of small sets of predictive features were proposed and applied to the analysis of genomic data sets. More precisely, the classification in molecular subtypes of tumor samples which had a variant of lymphoma cancer ("Diffuse Large-B Cell Lymphoma") was approached.

Autorentext

Andres Masegosa was born in 1980 in Cullar (Spain). He obtained a Master Degree (2003) and a Ph. Doctor Degree (2009) both in Computer Sciences at the University of Granada. Topics of interests include supervised classification models with both precise and imprecise probabilities and automatic and interactive learning of Bayesian networks.