Expand your toolkit
Keeping up with the growing ecosystem of machine learning algorithms has never been this easy. DataRobot’s vast set of diverse, best-in-class algorithms from R, Python, H20, Spark, TensorFlow, and other sources is constantly expanding and at your disposal. With one line of code or a single click, DataRobot trains, tests, and compares hundreds of different models, including techniques you may never have used from languages or projects you may not be familiar with.


Model optimization
Within every project, DataRobot automatically identifies the best pre-processing and feature engineering for each modeling technique. Employing text mining, variable type detection, encoding, imputation, scaling, transformation, and automated feature engineering (to name a few), DataRobot scientifically selects the hyper-parameters and options that optimize out-of-sample performance (for example, AUC, LogLoss, Gini etc.) for your models. Of course, the DataRobot platform also provides you with options to select your own tuning parameters and do custom feature engineering.
Massively parallel
DataRobot uses hundreds, or even thousands, of servers — as well as multiple cores within each server — to parallelize data exploration, model building, and hyper-parameter tuning. Large datasets? No problem. DataRobot uses distributed algorithms from Spark and H20 to spread the computation across your servers. The speed of building, deploying, and making your predictions is limited only by the computational resources at DataRobot’s disposal.


Easy model deployment
With DataRobot’s API, you can operationalize your models with just a few lines of code, regardless of whether you need real-time predictions, batch deployments, or scoring on Hadoop. Update your models with no downtime and no need to write new scoring code…ever.