Improving Time Series Prediction With Feature Selection Using a Velocity-Enhanced Whale Optimization Algorithm

Abstract

The nonlinearity and uncertain behavior of many recent financial applications is increasing rapidly. Thus, it is important to resolve the rapid growth of time-variant problems with the help of artificial intelligence methods. In this paper, a hybridized method is used to predict four types of financial datasets: absenteeism at work, blog feedback data, currency exchange rate, and energy consumption. The prediction accuracy is improved with feature selection techniques. During the use of feature selection methods, only related features are carefully chosen and then fed to the neural network algorithm for prediction. In this research, the previous year data is taken for training and recent year data is taken for testing. Finally, the results of the velocity enhanced whale optimization algorithm (VEWOA) is compared with other methods like local linear wavelet neural network (LLWNN) and local linear radial basis functional neural network (LLRBFNN).