decisionintelligence
Plot model predictions against target
Hi again !
From the csv generated following your 'steps_to_get_predict_and_actual_data.md' file, how do you manage to plot the predictions ? Do you have some chunk of codes I could reuse ?
Thanks in advance, Best, Alexis
Sorry, we currently do not have the code to draw the predicted results, but we are working on an EasyTime platform that allows for visualization. But it may take some time before users can officially use this platform.
EasyTime: Time Series Forecasting Made Easy
Ok thanks.
When I dig into the values and extract the predicted values for instance, I get a string representing nested lists :
[ 1.15842237e+01 2.60637379e+00 9.18402386e+00 1.10782933e+00 2.78504157e+00 1.09934306e+00 8.00008392e+00] [ 1.15357113e+01 2.74651909e+00 8.59091568e+00 1.18943906e+00 2.80523276e+00 1.07728708e+00 7.83176613e+00]
How have you arranged the results ? How would you extract them ?
This result is normal because rolling forestry predicts each window and stores the true and predicted values of each window accordingly. To parse these results, you can refer to steps_to_get_predict_1and_ctual_data.md, which provides a method.
I have parsed the results following the code you mentionned. Then how do you exploit the csv to get a simple visualization ?
You will receive the actual and predicted values for each window, and then you can write Python code to visualize them.
Ok, which goes back to my initial question : how do you manage to plot the predictions at this point ? Do you have some chunk of codes I could reuse ? It seems the actual and predicted values are encoded as a string.
Ok. Could you just explain how you arrange the data inside the csv ? The data seems to be arranged in a nested list in a string format like this :
[ [[7 values] [7 values] ..... [7 values]] [[7 values] [7 values] ..... [7 values]] .... ] (see screenshot)
I am studying the ETTh1 dataset with 96 points to predict. Where are the 96 context values and the groups of 96 predicted values here, corresponding to one window (so 192 points in total) ?
Because ETTh1 has 7 variables, you will see 7 variables together, and because you predict 96 steps, you will see 96 sets of 7 variables.
If there are no other questions, I will end this issue. If you have any questions, please feel free to open a new issue for communication!
Hi ! Sorry for the delayed response. Yes I managed to plot the figures. I have made a pre-print of an analysis of TSF models based on dynamics considerations. We use the code and results of your benchmark. Here is the link of the paper if you want to have a look at ! 😄 https://arxiv.org/abs/2507.15774
Thank you for your support and recognition. If you have any suggestions for improvements, feel free to share them. Excellent work—reading it was truly insightful. I wish you the best of luck with your submission！！！
Sure ! Plotting the predictions was quite painful ! So any possibilities to make it easier would be great. Thanks for reading it ! Same for you, if you have any feedback on our work, feel also free !