Global and decomposition evolutionary support vector machine approaches for time series forecasting

Abstract

Multi-step ahead Time Series Forecasting (TSF) is a key tool for supporting tactical decisions (e.g., planning resources). Recently, the support vector machine emerged as a natural solution for TSF due to its nonlinear learning capabilities. This paper presents two novel Evolutionary Support Vector Machine (ESVM) methods for multi-step TSF. Both methods are based on an Estimation Distribution Algorithm (EDA) search engine that automatically performs a simultaneous variable (number of inputs) and model (hyperparameters) selection. The Global ESVM (GESVM) uses all past patterns to fit the support vector machine, while the Decomposition ESVM (DESVM) separates the series into trended and stationary effects, using a distinct ESVM to forecast each effect and then summing both predictions into a single response. Several experiments were held, using six time series. The proposed approaches were analyzed under two criteria and compared against a recent Evolutionary Artificial Neural Network (EANN) and two classical forecasting methods, Holt-Winters and ARIMA. Overall, the DESVM and GESVM obtained competitive and high quality results. Furthermore, both ESVM approaches consume much less computational effort when compared with EANN.

Cascaded evolutionary algorithm for nonlinear system identification based on correlation functions and radial basis functions neural networks

“…In this case, the designer would feed the algorithm with input and output data and get the final model with the selected inputs, complexity and related parameters. From the soft computing community, the works by [29][30][31][32][33] serve as examples to take inspiration and apply to automatic model building for the purpose of dynamical system modeling.…”

Online suicide prevention through optimised text classification

“…• For non-linear kernels, we varied the free parameter γ between 2 −14 and 2 4 (stepping by a factor of 4), and the polynomial degree d between 2…”

“…They have been shown to work well for jointly estimating features and hyperparameters for SVM[4] and offer the advantage that optimisation is initialised from a variety of points in the search space. Evolutionary algorithms borrow the concepts of fitness-based selection, mutation, inheritance and evolution, and apply them to a search problem.…”

Multi-step time series prediction intervals using neuroevolution

“…Popular examples include the Holt-Winters method and the ARIMA methodology [1]. Soft computing approaches are more related to the fields of computer science and include a range of distinct methods, such as [4,5]: neural networks (NNs), fuzzy techniques, support vector machines, evolutionary computation and even hybrid combinations of the previous methods. In particular, several works have used evolutionary computation to successfully optimize NN for single-point TSF [6,7], in a hybrid combination that is known as neuroevolution [8].…”