I have a couple of time series related questions. Please take into consideration that I have no experience about time series prediction :)
Let's say, I wish to predict what might be the value of "gold" two months later from today. However, I have a prerequisite: I wish to predict what might be the value of "gold" two months later from today, assuming the euro/dollar exchange rate becomes 1.05 at that time. In other words, I need to link the prediction of gold to euro/dollar rate with the assumption of euro/dollar rate will take a particular value (1.05).
This is not my area of expertise, but can use the lag columns node to create new features that look back in time and use these to training your models. If you want to "game" these models using different currency exchange rates, I would use a double input quickform node and variable to table column to create this additonal column in your data set.
it seems pretty straigh forward for time series from t-n until t (being the last data in the series), but the problem is how to generate values for t+1, t+2, t+3, ... meaning forecasting the data after the end of series.
I did this once but we never published it. The basic idea is, I predict one step and feed it back using the recursive loop so it will be your next input time step for the predictor. I attach you a workflow.
Dear Iris:
I'm doing a model to predict a time series . The model is made with an artificial
neural network, it has 12 lags and some variables of momentum and step on month.
The problem arises when I make a prediction, because of the lags as from the second month
of prediction, I have to include the prediction that the same model has given me.
I was trying to use your workflow, but I've noticed that you're using both modeling to
predict at the same time, and that differs from mine, since I have a node where I read the
model done and precisely adjusted to make predictions 1 year in the future.
I was trying to use your flow, but unfortunately could not.
I would be grateful if you can tell me how. :-)
I enclose my workflow prediction model that includes your workflow with recursive loop.
Hello, I am relatively new to knime and really new to time series analyses in Knime. Would anyone know how to access an online course for building time series prediction (specifically for predicting exchange rates or commodity prices)?
Hello, I am relatively new to knime and really new to time series analyses in Knime. Would you know how to access an online course for building time series prediction (specifically for predicting exchange rates or commodity prices)? Do know any way of a standard model or something that may use for grasping it?