A Hybrid System Based on Dynamic Selection for Time Series Forecasting

Abstract:

Hybrid systems, which combine statistical and machine learning (ML) techniques using residual (error forecasting) modeling, have been highlighted in the literature for their accuracy and ability to forecast time series with different characteristics. Due to random fluctuations, complex nonlinear patterns, and heteroscedastic behavior, residual modeling is crucial in these architectures. Thus, selecting, specifying, and training an ML model to forecast residuals is costly and difficult because underfitting, overfitting, and misspecification can result in a system with low accuracy or even deteriorate the linear time series forecast. This article proposes a hybrid system, dynamic residual forecasting (DReF), that uses a modified dynamic selection (DS) algorithm to choose the best ML model to forecast a residual series pattern and if it is a promising candidate to improve the linear combination’s time series forecast. The DReF reduces ML model selection uncertainty and time series forecast degradation. The system also finds the best DS algorithm parameters for each data set. The proposed method uses multilayer perceptron, support vector regression, radial basis function, long short-term memory, and convolutional neural network, five widely used ML models. Ten famous time series were tested. The DReF outperforms literature single and hybrid models for most data sets.

Note: Please discuss with our team before submitting this abstract to the college. This Abstract or Synopsis varies based on student project requirements.