Our AI writing assistant, WriteUp, can assist you in easily writing any text. Click here to experience its capabilities.

Time Series Forecasting Using Past and Future External Data with Darts

Summary

This article discusses how Darts, an open source Python library, can be used to easily take “covariates”, or other time series providing useful information, into account when forecasting time series. The article explains the distinction between past and future covariates and how Darts can be used with each. It then walks through an example of forecasting a river flow using synthetic time series data. The article also explains how to evaluate the different models and compares the results from using no covariates to using past and future covariates.

Q&As

What is Darts and what is its primary goal?
Darts is an open source Python library whose primary goal is to smoothen the time series forecasting experience in Python.

What are the two types of time series used for forecasting?
The two types of time series used for forecasting are past covariates and future covariates.

What is a toy example used to demonstrate how covariates can be used?
The toy example used to demonstrate how covariates can be used is forecasting a river flow.

What is the impact of past and future covariates on time series forecasting?
The impact of past and future covariates on time series forecasting is that they can help to improve the accuracy of predictions.

How can Darts be used to easily take "covariates" into account?
Darts can be used to easily take "covariates" into account by providing the past_covariates or future_covariates arguments to the fit() and predict() methods of the models.

AI Comments

👍 This article provides a great overview of how Darts can be used to easily take into account past and future covariates when forecasting time series.

👎 The example used in the article is simplistic and does not accurately represent how to forecast an actual river flow.

AI Discussion

Me: It's about how to use past and future external data to do time series forecasting using Darts. It explains the distinction between "past" and "future" covariates and how Darts can be used to take them into account when building models.

Friend: Interesting. What are the implications of this article?

Me: It's a helpful guide to using Darts for time series forecasting. It demonstrates how to use both past and future covariates, and how using them can improve the accuracy of forecasts. In addition, it explains how Darts can make the experience of using covariates easier and less error-prone.

Action items

Technical terms

Past Covariates
Time series whose past values are known at prediction time.
Future Covariates
Time series whose future values are known at prediction time.
BlockRNNModel
A model that supports past covariates and can be used to make predictions.
NBEATSModel
A model that supports past covariates and can be used to make predictions.
TCNModel
A model that supports past covariates and can be used to make predictions.
TransformerModel
A model that supports past covariates and can be used to make predictions.
RegressionModel
A model that supports both past and future covariates and can be used to make predictions.
LinearRegressionModel
A model that supports both past and future covariates and can be used to make predictions.
RandomForest
A model that supports both past and future covariates and can be used to make predictions.
RNNModel
A model that supports future covariates and can be used to make predictions.
ARIMA
A model that supports future covariates and can be used to make predictions.
VARIMA
A model that supports future covariates and can be used to make predictions.
AutoARIMA
A model that supports future covariates and can be used to make predictions.

Similar articles

0.82498693 Accessibility Links

0.8147965 AI for execs: How to cut through the noise and deliver results

0.81285757 Zero-ETL, ChatGPT, And The Future of Data Engineering

0.80582184 Transformers: the Google scientists who pioneered an AI revolution

0.8056732 Banks are falling behind when it comes to understanding — and using — data

🗳️ Do you like the summary? Please join our survey and vote on new features!