Estimación de las propiedades de curado de gomas mediante SVM

Resum

The Support Vector Machine (SVM) is a novel type of learning machine, based on statistical learning theory, with the capability of learning separating functions in pattern recognition (classification) or performing functional estimation in regression problems. The concept of SVM was introduced by V. Vapnik in the late 1970’s in pattern recognition problems. In the 1990’s the method was generalized and nowadays a growing interest has emerged as a result of the remarkable efficiency shown by SVMs, especially when compared with traditional artificial neural networks, like the multilayer perceptron. The main advantage of SVM, with respect to neural networks, consists in the structure of the learning algorithm, characterized by the resolution of a constrained quadratic programming problem, where the drawback of local minima is completely avoided. In this paper, we present an SVM model applied to the particular case of predicting the cure characteristics of the blends according to their chemical composition and mixing conditions. The advantages of this methodology are showed by comparing the obtained results with those obtained by mean of a feedforward neural network model, trained with the backpropagation algorithm, proposed in a previous work.