C4W4_DeepLearning.AI TensorFlow Developer Specialization

Course Q&A TensorFlow Developer Professional Sequences, Time Series and Prediction
1

Hello everybody,

on the last assignment of course4/week 4 in [DeepLearning.AI TensorFlow Developer Specialization] we are building a model to forecast the dataset “daily-min-temperatures.csv”.

The dataset hat a total of 3650 datapoints and in the excercise we are predicting and evaluate the points from 2500 to 3650.

I was wondering how I need to adjust the code if I want to predict further datapoints. For example 3651-4000. Of course I can not evaluate then, but I would like to forecast it.

Can anybody help?

Thank you ver much.

2

It seems like you are asking how to predict multiple points into the future.
Let’s call the number of points you want to predict into the future as F
Please adjust your the following functions in the notebook:

  1. Function windowed_dataset should return the inputs and outputs correctly. In your case, it’ll be a F number of points for prediction.
  2. The final layer of the model architecture should output F time steps instead of 1 time step.