0
$\begingroup$

I'm trying to use the Tensorflow regression tutorial (with Keras) to do some regression on a time series with a couple of inputs. I will provide code if asked so.

My inputs are: [Day, Hour, Minute, Soil Temperature, Air Temperature, Soil Moisture, Air Humidity]

My output or value I want to predict is: [Air Temperature]

My data consists of samples of every 10 minutes and I thought using these inputs I can predict the Air Temperature using this type of (feed forward) neural network as in the tutorial.

So what I am doing is feeding values at time T and predicting T+1

My configuration is:

Neural Network Configuration I am training the network on a 20 days worth of data and trying to predict the next 10 unseen days but my output looks shifted to the right by 1 and I cannot figure out why. I have even tried predicting my train data and I get the same story.

This is how the first couple of data-points for my prediction versus true value looks like: Output

So overall the prediction is not too bad but it appears to be shifted for some reasons.

$\endgroup$
0
$\begingroup$

Ok so now I understand that this is reasonable using a simple NN so now I'm trying a LSTM NN and still I get this shift. What's the best way to go about this then?

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.