A Hybrid Semi-linear System for Time Series Forecasting

Time series forecasting is a challenging task in machine learning. Each time series may be composed by linear or nonlinear patterns which need to be mapped by techniques such as autoregressive integrated moving average (ARIMA) and artificial neural networks (ANN). This work proposes an evolutionary semi-linear artificial network for time series forecasting. The system selects the best architecture for linear and nonlinear components of the ANN in order to deal with different patterns simultaneously. Particle swarm optimization is used to find suitable architecture and weights. Experiments show that the proposed technique achieved promising results in time series forecasting.