Perform Automated Time Series Modeling with DataRobot’s Latest Tool
Overview
- DataRobot has launched a platform that automates time series modeling
- The tool automates tasks like featuring engineering, detecting stationarity and seasonality, among others
- Core time series models like ARIMA and Facebook Prophet are supported, along with advanced tree based models
Introduction
Time Series modeling is one of the most complex and tricky tasks in data science. There are so many facets and intricate details a data scientist needs to consider before getting to the model building stage.
The end goal of building a time series model is to predict future performance based on historical data that is time-dependent. This includes examples such as predicting the sales of a restaurant in a holiday season, or the rise and fall of stock prices, etc. It might sound easy to read, but once you look at the data and the noise within it, the complexity goes to an entirely new level.
Data Robot, a leading company in the automated machine learning (AutoML) space, has taken all these things into consideration and designed a platform that automates the entire time series modeling process. Their aim is to help business leaders improve forecasts for sales volumes, product demands, staffing, inventory and whole lot more.
There are a lot of tasks this tool automates, including:
- Feature engineering
- Detecting stationarity
- Detecting seasonality
- Transforming the target variable
- Implementing backtesting to achieve the most accuracy possible
The core time series models like ARIMA and Facebook Prophet are present and are supported by advanced models including eXtreme gradient boosting models. The tool also offers full API support to integrate your models into the existing business processes and products.
DataRobot has also released a short video about the tool which you can view here.
Our take on this
This release, as with all automated machine learning platform releases, will help democratize machine learning and take the technical aspect out of the equation (as much as possible). AutoML is seeing a massive spike in interest with multiple tool releases lately.
I particularly like this release because I have seen even experienced data scientists get stuck with time series problems. This tool, if your business can afford it, will be a huge boost.
Data scientists worried about their jobs with AutoML don’t need to worry. These tools have been built to assist, rather than replace, existing jobs. It allows you to focus on the problem rather than the programming part. You can still tune and perform hyperparameter optimization, etc. but the tool does that for you to quite an extent.
Subscribe to AVBytes here to get regular data science, machine learning and AI updates in your inbox!
Senior Editor at Analytics Vidhya. Data visualization practitioner who loves reading and delving deeper into the data science and machine learning arts. Always looking for new ways to improve processes using ML and AI.
