KNIME Review

Our average record size was around 10 million records. If we have bigger data, we can opt for a Big Data extension for Hadoop, Spark, etc.


What is our primary use case?

In demand forecasting projects to extract, to clean and to transform data from various resources. Also some clustering and classification techniques are used for behavioural clustering and classification according to attributes.

How has it helped my organization?

My organization's field of activity is to develop business applications for niche areas. Almost three years ago, we decided to extend our solutions with advanced analytics. KNIME let us start easy and fast into the Advanced Analytics area. We are able to try project ideas with KNIME by doing proof of concept easy and prototyping fast.

What is most valuable?

  • Easy ETL operations
  • Rich algorithm set
  • Integrated with other languages like R, Python, and Java.
  • Works together with other technologies like DeepLearning4j, H2O.ai, D3.js, and Weka.
  • Ease of use and being a performant solution.
  • Continues development and wide community support

What needs improvement?

I mentioned about the distributed architecture in my previous answer, but they did with version 3.5. This time maybe I could add the integration with graph databases like Neo4j.

For how long have I used the solution?

One to three years.

What do I think about the stability of the solution?

In the previous versions, I had some issues when reading large Excel files due to memory usage. But with the previous version (3.3), they renewed all Excel nodes and now I have no problem. 

What do I think about the scalability of the solution?

With the data sizes that I dealt with, I did not. Our average record size was around 10 million records. If we have bigger data, we can opt for a Big Data extension for Hadoop, Spark, etc.

How is customer service and technical support?

I used it very little. All of them replied to me in one day. (It was not professional support, just over a forum). Also, I can find enough information in the documentation and forum.

Which solutions did we use previously?

Before KNIME, I used SQL language and Excel for data analysis but machine learning algorithms. In parallel to KNIME, I worked on a few projects with R and Python separately. So I cannot say that I switched from different solutions.

But just for ETL with Excel, KNIME brings me better visualization, rich function set, preserving operations to repeat again and better performance on the same hardware.

How was the initial setup?

I am using Mac and it is so easy. Download a .dmg, extract it as an app, and copy it to the applications folder. On windows it is also simple installation.

For extensions like R or Python, you need experience with general OS and installation processes.

What about the implementation team?

We did in-house.

What was our ROI?

The biggest ROI comes from productivity when creating new things and also supporting old jobs.

And there is no hidden cost. Licensing is simple and open than other platforms.

What's my experience with pricing, setup cost, and licensing?

KNIME is open sourced platform and has a free desktop version with unlimited data size and functionality.

Also, the server version is good for teams and enterprise productivity. Especially the new "Model Factory", which lets data science teams easily build and manage models. When compared with similar products, it is less expensive but as powerful as (or maybe more powerful than) others.

Which other solutions did I evaluate?

The Open Source licensing and community support is one of our important criteria. The second one is the interoperability with other technologies and openness to different data sources. There are two options: RapidMiner and KNIME. We chose KNIME.

What other advice do I have?

Data Science requires freedom for creativity. Sometimes you need to crawl data from the web or social media. Sometimes you need to blend different sources like NoSQL MongoDB and Excel files, etc. It is not only algorithms and data extraction, visualization and preparation steps are important as at least algorithms.

Don't go with software that has complex and hidden licensing costs, which will kill your flexibilty and creativity. Also, interoperability brings the advantage of limitlessness.

Disclosure: My company has a business relationship with this vendor other than being a customer: Partnership with KNIME
1 visitor found this review helpful
Add a Comment
Guest
Sign Up with Email