LSTMs vs VAR models

VAR models (vector autoregressive models) are used for multivariate time series, especially in the field of macroeconomics. In this article, we are going to check if LSTMs (Long short-term memory neural networks) can outperform VAR model predictions. The VAR model is going to be estimated in R, while the LSTM is created in Python. In our models, we will be using four variables, all of which belong to quarterly Spanish economical data. First, we preprocessed the data to unify and compare information from different sources (you can find how in the R script).