Abstract: In the realm of time series forecasting, traditional models often struggle to effectively manage the intricate dependencies and high dimensionality inherent in multivariate data samples.