What models can we use if we have to forecast the sales of eighteen products for a full year? In this video, we're going to learn NBEATSx, which is an MLP-based deep learning model for long-term forecasting.
NBEATS and NHITS have been covered in my other videos, and this video completes the group. Even though there are many similarities, in this video I talk about the model's components and how we can train the models in more detail.
I also give my thoughts about the model's performance. Do you have any other opinion? Please share with us!
00:00 Intro
02:08 Model overview and comparison with NBEATS and NHITS
06:25 The 'x' in NBEATSx
06:52 The basis expansions
25:28 Why I use this dataset?
27:21 The model performance
30:38 Explore the data
39:27 Prepare the data
40:59 Train the models
48:37 The model performance on the validation set
52:32 Train the models using all of the data
53:15 Discussion
Source code: https://www.kaggle.com/code/leesstephanie/nbeatsx-for-kaggle-stores
Thank you so much for watching!
#timeseries #forecasting #deeplearning #nbeatsx #kaggle #salesforecasting #neuralnetworks