What's New in KNIME Analytics Platform 3.2, KNIME Server 4.3 and KNIME Big Data Extensions 1.6

UI and Workbench

KNIME File Extension

We now have dedicated file extensions for Workflows and Workflow groups.

• *.knwf for KNIME Workflow Files
• *.knar for KNIME Archive Files

Once you have registered these file extensions you can open them by double-click, which directly launches your KNIME Analytics Platform. You can now also Drag&Drop the files to the KNIME Explorer to import them into your repository.

Missing Node installation

It is no longer necessary to search for the correct plugin for a node yourself. The new version provides an automatic Missing Node Installation. Now, when you open the workflow containing missing nodes, the dialog (as shown here) tells you which plugin is missing and, if you then select it, will directly search in your active update sites for the respective plugin.

Workflow Coach

The Workflow Coach is our brand new node recommendation engine. Based on our communities' usage statistics we can now give you hints as to which node to use next in your workflow. When applied together with our Personal Productivity Extension (which requires a purchased license) you can even use this based on your own workflow. And for our KNIME Server customers – you can also make those statistics available to your users individually.

Wrapped Metanodes

A small change with high impact: Open the context menu for a set of selected nodes and have them directly encapsulated as a Wrapped Metanode. This saves two steps when designing workflows intended for use in the KNIME WebPortal or the Streaming Executor.

Streaming

We have further improved our Simple Streaming Executor and converted more nodes to be able to support the new streaming API – including popular nodes such as String Manipulation and all of the Rule Engine nodes.
What is more, the Streaming Executor is now also able to process nodes that require additional handling such as extra scans over the data before main execution. You can find more information about the Streaming
Executor on the KNIME Blog.

Analytics and ETL

Feature Selection

The new Feature Selection nodes can be used to identify important features and reduce the dimensionality of your data. The supported selection strategies are forward selection and backward elimination. You can also specify your own score variable to optimize, which provides greater flexibility. Alongside the nodes there are two preconfigured meta nodes, one for each of the selection strategies.

Ensembles of Trees and Gradient Boosted Trees

Gradient Boosted Trees – Alongside Random Forests, the new KNIME Analytics Platform includes nodes for Gradient Boosted Trees, which are a specialized version of Gradient Boosting. They are considered to be among the state-of-the-art solutions for classification and regression problems. This new set of nodes includes nodes for learning and predicting both classification and regression problems.

Enhancements for Random Forests and related nodes – A number of enhancements has been added to the Random Forest related nodes (Tree Ensemble, Gradient Boosted Trees, Simple Regression Tree and Random Forest). Most notable is the optimization of the tree building algorithm, which now enables Random Forests to learn much faster on large data sets. We have also added binary splits, which allow for more interaction between the different features in a data set and oftentimes provide better generalization properties.
Furthermore all nodes in the Decision Tree Ensemble category, as well as the Simple Regression Tree, are now able to handle missing values in the data, which was not possible before.
One last enhancement involves the translation of Random Forest regression models as well as Simple Regression Trees to PMML.

Deep Learning

Deep Learning – The KNIME Deeplearning4J Integration in KNIME Labs (developed with the Deeplearning4J library) extends KNIME Analytics Platform providing the functionality to use deep neural networks. The extension consists of a set of new nodes, which enable you to modularly assemble a deep neural network architecture, train the network on data, and then use the trained network for predictions.

Neural Word Embeddings – In addition to deep learning, the KNIME Deeplerning4J integration contains nodes to learn word embeddings from words and documents. This is accomplished using a Word Vector Learner Node. This can create meaningful numerical representations of text words that can be used in many applications.

PMML Transformation Applier

Integration and Utility Nodes

REST Service Client Nodes

Part of KNIME Labs and following up on the famous KREST community extension, KNIME Analytics Platform now provides a set of REST client nodes to integrate RESTful web services. These nodes provide a rich configuration to reach out to different services and are tightly integrated in KNIME's JSON and XML processing capabilities.

Tableau Integration​

KNIME Tableau integration is now available as part of KNIME Labs. Tableau is a powerful business intelligence solution to build highly interactive and powerful dashboards. The integration comprises two new nodes:
Tableau Writer (TDE) writes a KNIME table as a Tableau extract (TDE); and Send to Tableau Server exports the table directly to a Tableau server for further processing. Special thanks to Tableau for their assistance during development and Forest Grove Technology, our partner in Australia, for valuable feedback and testing.

Semantic Web

Have you ever wondered how Google answers queries like “When was Albert Einstein born?” These queries are powered by the Semantic Web, the Google Knowledge Graph. The new Semantic Web nodes in KNIME Labs enable access to these semantic resources, e.g DBpedia or CHEMBL, from within KNIME.

JavaScript Views

Parameterized Database Query

H2 Database Connector

KNIME Big Data Connectors and KNIME Spark Executor
(commercial extensions)

KNIME Big Data Connectors

KNIME Big Data Connectors have been extended by two new remote file handling nodes that can access files in HDFS. The webHDFS Connection node can connect to HDFS via the webHDFS protocol, which requires network access to the HDFS NameNode and all DataNodes. The httpFS Connection node can access HDFS files via an httpFS gateway, that is installed on a cluster edge node as a single point of entry.

Multiple Apache Spark Versions Supported

KNIME Spark Executor v1.6 now allows you to run workflows with Spark nodes on clusters with Spark 1.5 and 1.6. This includes support for Hortonworks HDP (2.2, 2.3.0, 2.3.4, 2.4) and Cloudera CDH (5.3 - 5.7). You no longer need to choose the Spark version to connect to during installation, instead, a revised preference page makes it easy to switch between Spark versions.

KNIME in the Cloud and KNIME Server
(commercial extension)

KNIME Cloud Analytics Platform (on Microsoft Azure)

Get started quickly – find KNIME Analytics Platform on the Azure marketplace and launch a new machine in a matter of minutes.

Bring your analytics to cloud hosted data – KNIME Cloud Analytics Platform can be in the same data center as your data - so no more waiting when you transfer data from the cloud to your local machine.

Scale your Analytics – Azure offers machines with up to 32 cores and 448 GB main memory, far more than your laptop! Simply launch the machine suitable for your workload, and when you’re done shut it down to keep your costs under control.

See KNIME Cloud Analytics Platform in action.

KNIME Server License

KNIME Server Admin Portal

KNIME Server Installer

KNIME Server Rest API

The KNIME Server REST API has been further extended to allow upload/download of files and workflows. The following blog posts describe the REST API and its use.