timetk 2.0.0: Visualize Time Series Data in 1-Line of Code

Written by Matt Dancho on June 5, 2020



I’m ECSTATIC 😃 to announce the release of timetk version 2.0.0. This is a monumental release that significantly expands the functionality of timetk to go WAY BEYOND the original goals of the package. Now, the package is a full-featured time series visualization, data wrangling, and preprocessing toolkit.


Forecasting and Time Series Software Announcements

Articles on the modeltime and timetk forecasting and time series ecosystem.


Like these articles?
👉 Register to stay in the know 👈
on new cutting-edge R software like modeltime.


Background

I’ll go into detail about what you can do in 1-line of code, and I have several more articles coming on how to use several of the new features. But first a little history…

Time series has been a passion project for me since my days of forecasting sales and economic data for a manufacturing company I worked for. My one gripe has always been that I had to use 50 different packages (zoo, xts, dplyr, etc) made by 50 different people to perform common data wrangling and visualization analyses. timetk solves this problem by making a consistent approach to visualize, wrangle, and preprocess time series data inside the tidyverse and tidymodels ecosystem.

With advancements in tidy-time series, the combination of the tidyverse and time series is an amazingly powerful concept. I’m not the first one to think of this idea. In fact, Davis Vaughan created tibbletime and the “tidyverts” (Rob Hyndman, Earo Wang, and Mitchell O’Hara-Wild) have created a whole forecasting and data wrangling system using a tsibble data structure.

These are amazing packages, but they solve different needs. tibbletime focused on data wrangling. The tidyverts focused on forecasting at scale using ARIMA and company.

My needs are different. I need:

  • Interactive visualizations for easy data exploration
  • Time series data wrangling for doing time series summarization, filtering, padding, and simple date-based arithmetic
  • Transformations and preprocessing that fit into the NEW tidymodels ecosystem (so I can do Time Series Machine Learning in addition to ARIMA forecasting)

So I created timetk version 2.0.0 to solve these needs.

Mission

Here’s the new mission, and what you can do in 1-line of code with timetk >= 2.0.0:

To make it easy to visualize, wrangle and preprocess time series data for forecasting and machine learning prediction.

What Can You Do in 1-Line of Code?

First step, load these R packages.

library(tidyverse)
library(lubridate)
library(timetk)

Investigate a time series…

This is fun! In 1 line of code we can visualize a dataset.

# Static ggplot
taylor_30_min %>%
    plot_time_series(date, value, .color_var = week(date),
                     .interactive = FALSE, .color_lab = "Week")

plot of chunk unnamed-chunk-2

So what did we just do?

We are exploring taylor_30_min, which is a classic electricity-demand time series that comes from one of my all-time-favorite packages, forecast. It’s been updated to the tibble structure with a time stamp column called “date” and a value column called “value”.

We can plotted it using plot_time_series(). I set .interactive = FALSE to return a ggplot (I’ll do that for all of the visualizations in this tutorial). The default is to return an interactive plotly graph, which is great for shiny apps, rmarkdown HTML documents, and super powerful for exploring time series (zooming, panning, etc).

Want the an interactive plotly visualization?

Just try this code and explore the plotly visualization.

# INTERACTIVE Plotly
taylor_30_min %>%
    plot_time_series(date, value, .color_var = week(date),
                     .interactive = TRUE, .plotly_slider = TRUE, .color_lab = "Week")

Let’s pick up the pace. Here’s some more amazing visualization capabilities!

Visualize anomalies…

We can visualize anomalies for multiple time series groups. Here we use group_by() to group the time series. Note this is a different dataset, walmart_sales_weekly.

walmart_sales_weekly %>%
    group_by(Store, Dept) %>%
    plot_anomaly_diagnostics(Date, Weekly_Sales, 
                             .facet_ncol = 3, .interactive = FALSE)

plot of chunk unnamed-chunk-4

Make a seasonality plot…

We can get seasonality plots.

taylor_30_min %>%
    plot_seasonal_diagnostics(date, value, .interactive = FALSE)

plot of chunk unnamed-chunk-5

Inspect autocorrelation, partial autocorrelation (and cross correlations too)…

And we can search the Autocorrelation and Partial Autocorrelation.

taylor_30_min %>%
    plot_acf_diagnostics(date, value, .lags = "1 week", .interactive = FALSE)

plot of chunk unnamed-chunk-6

Documentation

I have several more tutorials and a full time series course coming. Visit the timetk website documentation for more information and a complete list of function references.

Advanced Time Series Course
Become the times series domain expert in your organization.

Make sure you’re notified when my new Advanced Time Series Forecasting in R course comes out. You’ll learn timetk and modeltime plus the most powerful time series forecasting techiniques available. Become the times series domain expert in your organization.

👉 Advanced Time Series Course.


You will learn:

  • Time Series Preprocessing, Noise Reduction, & Anomaly Detection
  • Feature Engineering using lagged variables & external regressors
  • Hyperparameter Tuning
  • Time Series Cross-Validation
  • Ensembling Multiple Machine Learning & Univariate Modeling Techniques (Competition Winner)
  • NEW - Deep Learning with GluonTS (Competition Winner)
  • and more.

Unlock the High-Performance Time Series Course

Have questions on using Timetk for time series?

Make a comment in the chat below. 👇

And, if you plan on using timetk for your business, it’s a no-brainer - Join my Time Series Course Waitlist (It’s coming, it’s really insane).