The Time Series Forecasting Playground
- Posted by Daitan Innovation Team
- On June 2, 2021
- AI, Open Source, Time Series Forecast
A new web-based tool to get insights on time series forecasting
Time Series forecasting comprises a set of algorithms that are designed to predict future behavior based on historical data. Here at Daitan, time series forecasting has been one of the most important applications of machine learning and today, we are pleased to announce the first (to our knowledge) time series forecasting playground. Inspired by the Neural Network Playground and the GAN Lab, the The Time Series Playground is an interactive open-source tool designed to provide intuition on how to train AutoRegressive Feed Forward Neural Networks for time series forecasting.
In the tool, one can define, configure, and train Neural Networks using four different time-series “toy” datasets. For each dataset, users can experiment with different kinds of input formats, and play with up to 5 different training hyperparameters, including the learning rate, the choice of activation function, the batch size and many more.
Also, one can start, pause, or resume training at any given moment; or even choose to train a model for a single epoch at a time. When a given training process is finished or paused, the tool automatically displays one-step-ahead forecasting for the test set in the main graph, along with the 95% confidence intervals. Besides the hyperparameters, users can define their own choices of train and test splits, or even customize the input data format to be used for training.
The time series playground is designed to be an educational tool. We hope that the tool can provide valuable insights and spark curiosity so that more people feel interested in dive deep in this interesting subarea of machine learning.
If you want to learn more about time series forecasting, we highly recommend our series of articles on the topic.