arXiv:2405.04539v1 Announce Kind: new
Summary: This analysis paper introduces revolutionary approaches for multivariate time collection forecasting based mostly on totally different variations of the mixed regression technique. We use particular knowledge preprocessing methods which makes a radical change within the behaviour of prediction. We examine the efficiency of the mannequin based mostly on two varieties of hyper-parameter tuning Bayesian optimisation (BO) and Traditional Grid search. Our proposed methodologies outperform all state-of-the-art comparative fashions. We illustrate the methodologies by means of eight time collection datasets from three classes: cryptocurrency, inventory index, and short-term load forecasting.
Supply hyperlink