Sequences, Time Series and Prediction: C4W1

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

When creating a post, please add:

  • Week # must be added in the tags option of the post.
  • Link to the classroom item you are referring to:
  • Description (include relevant info but please do not post solution code or your entire notebook):C4W1_Lab 2_Forecast
1 Like
2

hi i just wanna ask at this ungraded lab, at the smoothing section:
to find the diff moving avg + smooth from the past:

  1. when slicing why split_time -370 not 365?
  2. then split time -370:359,11 359 is the current year? since we gave to 370? so how we want to decide the windows time ? why is it 11 is it depend on our project?
1 Like
3

Have you posted in the right course, I dont think so?

4

is this the page?

5

Hello @fatihah
You have not posted this question under your course i.e. TensorFlow Developer Professional. You have posted under Generative AI

image
image887×81 8.5 KB

Using Edit option kindly recategorize into correct course. Then the respective mentor can address your query.

6

done understood now thanks for the guidance:)

7

Hi @fatihah

You also needed to move your query to the right section by using the edit pen option, for now I have moved it to the right category for response from the course mentors.

the split time is 370 as the slicing is related to the series (i.e. period and noise level of 5) and not only period.

If you check the generate synthetic data, the forecast is a seasonal trend of 365 steps but some noise is added as the real world data often has some noise.

In numerical analysis, noise is generated by any random process, such as due to measurement errors and environmental disturbances

Your next doubt

One technique you can use is to do a moving average. This sums up a series of time steps and the average will be the prediction for the next time step. For example, the average of the measurements at time steps 1 to 10 will be the forecast for time step 11, then the average for time steps 2 to 11 will be the forecast for time step 12, and so on.

The function below does the moving average for the entire series. It takes a window_size argument to indicate the number of time steps to consider when computing the mean.

So as the time steps here is taken from step 1 to 10 steps, step 11 is average of step 1- step 10, as the argument mentions
window_size (int) - the number of time steps to compute the average for

Also a general Rule of thumb: A commonly used rule of thumb for choosing the window size is to use 10% to 25% of the length of the time series . For example, if the time series has 100 data points, a window size of 10 to 25 data points might be appropriate.

Regards
DP

8

Thank you so much now i understand how to adjust

9

@fatihah

Next to your headline of this post
Sequence, Time Series and Predictions:C4W1 you must have noticed the emicon :writing_hand:t2:, which need to be clicked to make any changes to your post headline, the category per se the course you are having query from and additional tags of which week, so the mentor responding also find more prompt response to address issue rather than searching for the course or assignment is where from ? If a learner finds it difficult to do this, can atleast the mention the details of specialisation, course, assignment name and week number in the body of the post with their explanation of the issue.

For more better use of discourse community, refer the FAQ section about community guidelines and not to directly DM to mentors until they ask you to do so are mentioned there.

Regards
DP