Hi Blaine, I definitely agree with you. Concerning exogenous factors there are several strategies how to add the to machine learning model — or as an additional variable (that can be repeated on each time step it it has different frequency) or with multimodal learning framework. Concerning multivariate time series it’s relatively easy from applied point of view. And concerning unknown lags, you’re totally right, I haven’t encountered it in my practice yet, but it’s a great issue to think about, thanks!

Co-founder of consulting firm Neurons Lab and advisor to AI products builders. On Medium, I write about proven strategies for achieving ML technology leadership

Co-founder of consulting firm Neurons Lab and advisor to AI products builders. On Medium, I write about proven strategies for achieving ML technology leadership