Granular data regression with neural networks

摘要: Granular data offer an interesting vehicle of representing the available information in problems where uncertainty, inaccuracy, variability or, general, subjectivity have to be taken into account. In this paper, we deal with a particular type granules, namely interval-valued data. We propose multilayer perceptron (MLP) model input-output mappings. The proposed MLP comes weights and biases, is trained using genetic algorithm designed fit different levels granularity. modeling capabilities are illustrated by means its application both synthetic real world datasets.