H2O AI and KNIME
KNIME users have worked with H2O-3 inside KNIME since 2017, and a KNIME Hub extension adds H2O Driverless AI, our AutoML. Teams can build one visual workflow from data access and preparation through automated model training, then score H2O MOJO models in KNIME with drag-and-drop predictor nodes. The integration brings the H2O Predictive AI Platform to analysts without leaving their KNIME environment.
Partnership overview
What KNIME users get
The expanded KNIME integration brings H2O automated machine learning together with the self-service data preparation, blending and guided analytics of KNIME. The joint solution automates more of the data science life cycle, from data prep and AutoML to MLOps and continuous optimization. Customers of H2O and KNIME can:
- Get fast, reliable and flexible access to data, insights and predictions across their organization's data
- Build a single integrated visual data science workflow, from data preparation to AutoML to model production
- Bring H2O Driverless AI, our AutoML, to business analysts in KNIME Analytics Platform and KNIME Server, to simplify and speed up machine learning
- Productionize Driverless AI models (MOJOs) on the KNIME drag-and-drop canvas, with model management and monitoring
Accelerate ROI on AI initiatives
- Expand automation. Automate the end-to-end data science workflow.
- Simplify data preparation. Drag and drop data sources and transformation processors.
- Ease of use. Use Driverless AI quickly and intuitively in a KNIME workflow.
- Accelerate model training. Build better models in less time with Driverless AI.
- Deploy in KNIME. Deploy H2O-3 and Driverless AI models (MOJOs) into KNIME workflows.
How the integration works
KNIME users have worked with open-source H2O-3 in KNIME since 2017. They can now use Driverless AI too, through an extension on the KNIME Hub. Data scientists and analysts work on machine learning projects faster, using automation and modern computing power to finish in minutes or hours tasks that can take people months.
- Develop an integrated data science workflow in KNIME Analytics Platform, from data discovery and data preparation to production-ready predictive models
- Give business analysts automatic machine learning with Driverless AI, so more of them work as citizen data scientists
- Reduce model deployment times, with Driverless AI and KNIME Server managing the production deployment process reliably
In a KNIME workflow, Driverless AI provides automatic feature engineering, model validation, model tuning, model selection, machine learning interpretability, time series, NLP, computer vision, and automatic pipeline generation for model scoring. It addresses the needs of many use cases, for organizations in every industry.
Simplify and automate data preparation in KNIME
KNIME provides a rich set of data source connectors and data preparation nodes on a no-code drag-and-drop canvas. Data analysts, data engineers and data scientists quickly build flows to prepare, wrangle, clean, join and filter data and get it ready for machine learning. Once the data is prepared, it connects to H2O-3 or Driverless AI to build machine learning models on the same canvas.
Driverless AI AutoML in KNIME
H2O-3, and now Driverless AI, extend KNIME Analytics Platform and KNIME Server with automatic feature engineering and machine learning. Data scientists, analysts and business users develop trusted machine learning models from their existing KNIME environments and run the entire data science workflow in one place.
Make predictions in KNIME flows
KNIME builds machine learning production workflows that consume models trained with Driverless AI. H2O provides production-ready, low-latency models and pipelines in the MOJO deployment artifact. A MOJO (Model Object, Optimized) is a standalone, low-latency model object that embeds easily in a wide variety of production environments. KNIME provides H2O MOJO and Driverless AI MOJO Predictor nodes to score data within a KNIME workflow through a drag-and-drop interface.
Extended no-code automation
Preparing data for AI, selecting the right model, pushing it into production and continuously optimizing it typically takes many stakeholders and several tools. Parts of this process can be automated, but flexibility matters: teams need to pick the techniques that best fit their machine learning use case. The H2O and KNIME integration covers these challenges, raises data scientist productivity, reduces overall IT spend and delivers more accurate predictions.
Data analysts, data scientists (from novice to expert) and data engineers combine KNIME and H2O workflows for data access, preprocessing, machine learning, deployment and optimization, and automate any or all of it without changing environments. They can work visually and without code, with the flexibility to code when needed. Users visually build a workflow and use AutoML to train a model, deploy a REST service or an end-user application, or mix in code to find the best method for their specific problem.
Resources
- Blog: combining the power of KNIME and H2O in a single integrated workflow
- Webinar: accelerate and simplify the data science workflow with KNIME and H2O
Try it
Try KNIME Server with Driverless AI