I’m developing several multivariate time series forecasting models, however I’m unsure which technique should I use as a benchmark
I’ve read several resources that recommends benchmarking against simple methods like naïve method, or moving average… but none of these considers the multivariate nature of my data.
I’ve built a Winter holt model, used only the target variable…however, this way I’m treating the benchmark as univariate. is it fair to compare its performance with multivariate models ?