How to Predict Multiple Time Series At Once With Scikit-Learn in 10 Simple Steps
I have seen many data scientists think about approaching this problem by creating a single model for each product.
Although this is one of the possible solutions, it's not likely to be the best.
Here I will demonstrate how to train a single model to forecast multiple time series at the same time.
This technique usually creates powerful models that help teams win machine learning competitions and can be used in your project.
And you don’t need deep learning models to do that!
We will use a sales forecasting dataset to learn, but this is a general method.
I have applied the same method to model both sub-minute market tick data and monthly sales predictions.
The 10 Steps You'll Learn
- Melt the data, stack the series
- Split the data into past and future
- Set up a 1-step target
- Compute 4 fundamental features
- Choose an evaluation metric
- Establish baseline
- Train the model
- Evaluate the model
- Extend the model to predict n-steps
- Predicting new examples
These 10 steps are split into 4 recorded video lessons of 10-13 minutes each.
So you can learn everything in an hour.
The code is available for download as a Jupyter Notebook and the data is linked as a small CSV.
100% Risk-free
If you are not 100% satisfied with your purchase, send an e-mail to support@mariofilho.com within 30 days and you will get a full refund without any questions asked.
Who is Mario Filho?
I am working with machine learning since 2014.
I am mostly known by being a Kaggle Competitions Grandmaster and prize winner in multiple competitions (on Kaggle and other platforms).
I have worked mostly as a freelancer, but was a Lead Data Scientist at Upwork Inc from December 2017 until early 2019.
And to finish it up, did more than 100 machine learning mentoring sessions with 5 stars at Codementor.io. (Yep, I was younger in the picture below)