KNIME Analytics Platform

KNIME Analytics Platform is a relatively new business analytics and data science tool with a visual workflow editor that lets professors and students focus on teaching and learning analytics concepts, methods and best practices. It is open-source and free with full functionality – no limitations on the number of students, methods, data or projects (education, research or professional). It is operating system agnostic – it natively runs on Windows, macOS and Linux platforms [1] – so students can keep it throughout their education and career and use it for any learning or professional engagement.  

Based on personal experience with a multitude of analytics tools available in the market, I can attest that KNIME offers one of the most comprehensive data access and data processing capabilities, modeling building and testing functions for a wide range of machine learning algorithms, data visualization and deployment options. An initial installation of KNIME comes with more than 2,000 native nodes (functions embedded in visual/graphical icons) to do almost everything a data scientist needs to do in analytics projects, although most common analytics tasks can be achieved with less than 30 nodes [2]. If these native nodes are not enough, one can build his/her own node, or use some of the nodes built by the community members (KNIME Community Extensions), integrate with current scripting and programming languages (e.g., Python, R, JavaScript), and/or integrate with popular reporting tools (e.g., Tableau, PowerBI, TIBCO Spotfire). A strong and active community that provides support and freely available learning materials and example workflows for teaching is also a differentiator.  

The following briefly describes some of the most prevailing reasons for considering KNIME Analytics Platform as a potential tool for data science learning, teaching and practicing.  

• Ease of use: The tool has a graphical user interface with an intuitive drag-and-drop, workflow-type model building logic and functionality that makes the analytics platform very easy and intuitive to learn, teach and use. 
• Cost of ownership: KNIME is truly free, not a scaled-down, time-limited trial or community version. The fully functional version is free for everyone (educators, researchers and practitioners) for everything – learning, teaching and consulting. 
• Rich functionality: With more than 2,000 native nodes and many more from third parties (via extensions), KNIME offers perhaps the richest collection of data science functions from data wrangling/blending/preprocessing to model building/testing/deploying.  
• Platform agnostic: Natively runs on all three popular operating systems – Windows, macOS and Linux. 
• Open source: The source of the analytics platform is open to everyone for exploring and innovating.  
• Community support: Access to KNIME gurus, as well as the large and highly active community of users for help on questions, modeling hints and example workflows, is just a few clicks away on KNIME Forum.  
• Modularity: This allows for the creation of reusable analytics models at varying levels of granularity via the use of metanodes and components where several nodes are combined and reused for repeated data science tasks.  
• Connectivity: Connects to almost every data source (local, web-based or cloud) and consumes any and every data type (structured, semistructured, unstructured).  
• Expandability: The analytics platform allows for building and sharing your own nodes and use of nodes built and shared by others. You can also expand the functionality via JavaScript, Python and R integration nodes.  
• Deployability: Putting workflows in perpetual use via deployment to KNIME Server on the cloud (or on-premise) is a quick and straightforward process. 

Having used pretty much every workflow-type analytics platform, both commercial and open source, I find KNIME the best option for a combination of the above-listed reasons. Out of these, the two that made my job more effective have been the high level of functional granularity and worldwide community support (especially the academic community). Many popular analytics tools in the market encapsulate several tasks under a single node/function, presumably making model building easier by taking the initiative to make intermediate decisions on behalf of the end-user (implying “trust me, I know how it should be done and what I am doing”). However, such an aggregated functionality may not be well suited for analytics teaching and learning. Students need to know the lowest level of tasks and decisions to be made to produce the best possible results. They need the opportunity to make mistakes, bad decisions, see the undesirable outcomes, and then go back and identify and fix the errors and imperfections.  

Another key differentiator for KNIME is its worldwide community engagement. As a faculty member, when I need an answer to somewhat unusual questions related to analytics functions and their specific implementations in KNIME, I use the community connections via the Hub and Forums. Not only do I get verbal answers, but often I get an exemplary workflow addressing my question shared by one of the community members. As a new academic user of KNIME, one can also find invaluable teaching resources for a variety of analytics and data science course offerings on the KNIME website at https://knime.com/academic-alliance. 

References 

1. Berthold, M. R., Borgelt, C., Höppner, F., Klawonn, F. and Silipo, R., 2020, “Guide to Intelligent Data Science: How to Intelligently Make Use of Real Data,” New York: Springer International Publishing. 
2. Silipo, R., 2020, “Why KNIME?,” Medium, July 14, https://medium.com/swlh/why-knime-98c835afc186. 
3. Delen, D. and Zolbanin, H. M., 2018, “The Analytics Paradigm in Business Research,” Journal of Business Research, Vol. 90, pp. 186-195.