Everyone is invited to a statistics talk by Professor Dr. Warsono from Indonesia:

"Classical Statistics and Deep Learning Hybrid Models in Forecasting Time Series Data"

Abstract: The main objective of this study is to promote a novel hybrid model that synergistically combines classical statistical methods and deep learning techniques to enhance the forecasting accuracy of time series data, especially for multivariate time series data. It leverages the strengths of classical multivariate time series models, such as Vector Autoregression (VAR), in capturing linear relationships among variables, and the prowess of deep learning techniques, such as Long Short-Term Memory (LSTM) models, in modeling complex nonlinear patterns over time. Our proposed hybrid VAR-LSTM model operates in two stages. Initially, the VAR component models the linear aspects of the time series. The predictions and residuals from this stage, which represent the nonlinear dynamics not captured by VAR, are then processed by the LSTM network. For illustration, this study presents the application of the proposed model in forecasting the multivariate time series data of Indonesia’s coffee export and import volumes.