We have been using the most recent version. It's version 4.6.

10 August 2023 - It has now been upgraded to 5.0 and is, if anything, even more impressive, especially in its ability to use Python and its libraries.

Knime seems to keep getting better. Their open-source model seems to be working. The addition of AI both to help in the building of workflows and as a facility within a workflow once it is up and running seems to add a dimension. At the moment, though, the system is so rich and fully featured that I have explored only the surface of the new version (5.4).

To date, all my needs have been met by earlier versions of Knime. I am, though, confident that should I need to start using version 5.4, the process will be smooth, and the new functionality fit for purpose. Upgrades to Knime have always worked like that in the past and I would expect them to do so in the future.

I used to be a Pascal programmer, and then I did a bit of Python. It does many of the things that I would've had to do in code, but does so without using code. I don't think it does everything, but it does most of what I need to do.

It can read many different file formats. It can very easily tidy up your data, deleting blank rows, and deleting rows where certain columns are missing. It allows you to make lots of changes internally, which you do using JavaScript to put in the conditional.

For example, I have one data set whereby all of the data is encoded and there was one variable called opinion or something like that and it had codes for what the topic was, which was being discussed, whether it was positive or negative, whether it was strongly worded or weakly worded, and so many other things like that.

I had to transfer those into columns, like sentiment, the strength of sentiments, topic being discussed. I had to split it up into columns, and I could do that very easily, like simple JavaScript, in their column expressions.

It also has very good fundamental machine learning. It has decision trees, linear regression, and neural nets. It has a lot of text mining facilities as well. It's fairly fully-featured.

They are also very careful with things like lab variants and issued variants because they have some labs that develop nodes, and new chunks of code which are represented as an icon. They make it very clear that those lab ones are not fully tested, and they're very glad to get comments back if you have problems.

I haven't had that difficulty myself. They seem to be aware that they have the community there as their testing base, and they seem not to be embarrassed about that. They will tell you when they go wrong and try to put it right.

So far, I haven't had problems with it, so I haven't really thought about room for improvement. It's so much better than many other things. It's useful in that you can at least get people who are pretty averse to programming to start thinking about putting something into a program of any kind, because they can see what's happening.

It's visual. It's codeless. For some purposes, I'd want to add Python or R, but I haven't had to do that so far, so I haven't seen the shortcomings of it. There must be some. All software has shortcomings, but I haven't recognized any myself.

Not just for KNIME, but generally for software and analyzing data, I would welcome facilities for analyzing different sorts of scale data like Likert scales, Thurstone scales, magnitude ratio scales, and Guttman scales, which I don't use myself.

I use both Thurstone scales and magnitude ratio scales quite a bit, and they're very powerful. But I've always had to do all the analysis myself in some simple code. I don't think that's provided. You could probably include it in KNIME, but I haven't tried to do it.

If it just said, "Analyze scales," and you'd choose which sort of scale you want to analyze and it gave you the options of normalizing or reversing or whatever it happens to be, that would be helpful. There are lots of simple functions that you want to apply to scales, which would be useful in any software, including KNIME.

I have been using this solution for about a year, but most particularly in the last six months.

It's been remarkably stable, much more so than most software. They have an active community forum. Problems seem to get fixed pretty quickly. I haven't had problems, but other people do report problems. So, there must be problems there, I just haven't had any.

On the very rare occasions that I have to seek advice, I just post it to the forum and someone will offer advice.

Positive

Compared to RapidMiner, at the moment I would go for KNIME, but that's largely because I haven't used RapidMiner much for the last year. It may have improved enormously since then. It was a very good package. They do much the same thing.

I'm more familiar with KNIME, so I would be able to talk more about it, whereas for RapidMiner, I was very enthusiastic when I used it. KNIME is a bit cheaper in a sense.

In RapidMiner, you can have up to 10,000 rows of data free of charge. For many things that I do, 10,000 rows of data is enough. I use quite a few UK government surveys, and I get the raw data from the UK Data Archive. They're often of the order of 10,000, 8,000. So, under 10,000 rows. I could use it free of charge.

I just downloaded it and then ran it. The process really was that simple. If I need one of the extensions (e.g. text mining), the process is just as simple.

We implemented the solution in-house.

I have not formally calculated it, but it must be substantial.

With KNIME, you can use the desktop version free of charge as much as you like. I've yet to hit its limits. If I did, I'd have to go to the server version, and for that you have to pay. Fortunately, I don't have to at the moment.

I would rate this solution an eight out of ten. 

I'm unwilling to give anything a ten because everything can be improved. But it's been very useful so far to me and has saved me many hours of work. I could have written it all in Python if necessary, but it would have taken me weeks for what would be a few days of work.

My advice is to just download it and use it. The documentation is pretty good. There are many good videos online for it. If you go to YouTube, you can get pages and pages of KNIME tutorials. They're pretty clear, and they are produced by people who've used it. It's not just company advertising, as far as I can see.