H2O.ai, the open source AI company, has released a new version of its automated machine learning platform H2O Driverless AI, with enhancements that it says will enable organisations to expand their current AI strategy and improve accuracy of predictive datasets.
The latest innovations in Driverless AI include time series support aimed at improving predictions within transactional datasets and auto documentation, as well as updated AutoViz, Machine Learning Interpretability (MLI) and Automatic Pipelines.
With the time series capability in Driverless AI, H2O.ai directly addresses some of the most pressing concerns of organisations across industries for use cases such as transactional data in capital markets, tracking in-store and online sales in retail, and using sensor data to improve supply chain or predictive maintenance in manufacturing.
“Time series is all-pervasive. It is not only one of the fastest growing classes of data but also one particularly well-suited for machine learning,” said Sri Ambati, CEO and founder at H2O.ai.
“Our latest recipes in Driverless AI automate predictions on time series for demand and inventory forecasting for retail stores, tick level predictions in capital markets and electronic trading, greener manufacturing by optimising supply chains and securing our world with sensor data from IoT.
“Our mission to democratise machine learning – to make it accessible to every enterprise unable to access the highly sought after talent in data science – is one step closer,” he added.
Driverless AI’s time series feature is said to be ideal for transaction, log and sensor data. Some of the key advantages include the ability to optimise for almost any prediction time window, whether that be for the next day or the next week, and incorporate data from numerous predictors, rather than focusing solely on past time series data.
It can handle gaps in time series data input without negatively impacting predictive modelling, handle missing values, structured character data and high-cardinality categorical variables as well as generate predictions for both numeric and classification problems.
In addition to the new time series support, H2O.ai has also added the following to H2O Driverless AI:
AutoDoc is an early-release feature of Driverless AI which generates a document describing the unique experiment pipeline chosen by Driverless AI and the user settings.
The report provides insight into the training data and any detected shifts in distribution, the validation schema selected, model parameter tuning, feature evolution and the final set of features chosen during the experiment.
Expanded AutoViz capabilities with additional charts and visualisations that enable data scientists to preview and examine a broader range of data sets.
Automatic Pipeline generation that provides the code needed to deploy scored models in applications on any device
data.table for Python is the data munging engine for Driverless AI and is H2O’s new project based on the popular open source project in R, data.table.
It brings the same features to the growing Python community and enables Python users to benefit from data.table’s high performance, superior data processing capabilities, and big data support
H2O Driverless AI empowers data scientists or data analysts to work on projects faster and more efficiently by using automation and state-of-the-art computing power to accomplish tasks that can take humans months in just minutes or hours.
It does so by delivering automatic feature engineering, model validation, model tuning, model selection and deployment, machine learning interpretability, time-series and automatic pipeline generation for model scoring.
