Deep Learning in Time Series Analysis

10 Hybrid Method for Cyclic Time Series

The context of time series analysis has received a large number of the modelbased learning methods for the purpose of classification in the last three decades. In contrast with knowledge-based methods in which information is extracted directly from the data, model-based methods presume a structure of the data flow in conjunction with a set of the hypothesis and parameters, whose values are found by the learning method. In fact, the outcome of a learning method is a set of values for the learning parameters, providing an optimal performance for the learning task, which is tightly dependent on training data. Hidden Markov model and artificial neural networks are known as the two model-based learning methods, ...