How can I predict the next month with time series?

Research & Models
, ,
#1

Hello everyone, I have a question related to time series. I have a year of data of a specific machine on its fuel consumption, I would like to predict an approximate total consumption for the next month (the value). Following the example of Tensorflow, the prediction works for me, but only for periods already known (checking the prediction). but I have modified the Window of time and I have not been able to predict a future date.

I am following this procedure

https://www.tensorflow.org/tutorials/structured_data/time_series

any recommendation or any idea? I really precited!

Thanks , Ricardo.

#3

Hey @Ricardo can you share the requirement that specifics about your needs, especially with the data?

Modeling time series issues are based on how you outline the samples and labels. There are basically two different ways roughly from high level views:

  1. Single step to predict
  2. Multi steps to predict.

This is my point of view.

Actually it is about how you want to choose the time window to train your model.

#5

Thanks for your answer! I have the predictions in a single step and also in multi steps, I am going to show you one of them.

Dataset

1
11262×907 86.3 KB

Time window

here some questions,

  • How can I see the date instead of time (h)?
  • For this problem is it a good idea to work with hours?
  • How can I see a date in the future? I only see dates that already have a label

2
21352×915 126 KB

Prediction

  • For example, if I wanted this prediction, how can I get the data as a value?

3
31347×1080 172 KB

Thank you so much, Ricardo.