“…The regression models include state-space approaches, such as hidden Markov models (MacDonald and Zucchini, 1997), generalized linear models (Green and Silverman, 1994), (McCullagh and Nelder, 1989) and logistic regression models (Liang and Zeger, 1989). Machine learning techniques, such as Support Vector Machines (SVM) and different types of neural networks, are also increasingly applied for time series regressions (Yadav et al, 2007), (Zhang and Wan, 2007), (Zainuddin and Pauline, 2011), also in the field of reliability prediction where they have proved to be powerful tools (Xu et al, 2003), (Pai, 2006), (Chatterjee and Bandopadhyay, 2012). However, the most commonly used neural network model, multilayer perceptrons (MLP), generally, suffer from drawbacks which include obtaining sub-optimal solutions due to local minima and requiring long computation time.…”