4

SPSS Modeler seems like a great tool for data mining (especially for prediction etc.) but it is extremely costly for individuals like me (around 20,000 euros excl. tax). There is also a video.

I am wondering if software as expensive as this really adds value. If it does, my university may consider purchasing it.

What seems unique with this software is that it allows you to do advanced computations without having the programming and statistical knowledge required for languages such as R.

What are your experiences with SPSS Modeler? How does it compare with other GUI based statistical software?

ttnphns
  • 51,648
  • 40
  • 253
  • 462
Tom
  • 469
  • 3
  • 9
  • 23
  • Tom, (only) Mods can turn post into CW. The simplest solution is to flag your post for mod attention so that we get notified. – chl Jan 08 '12 at 12:00
  • IBM SPSS Modeler, previously known as SPSS Clementine, is a full-scale **data mining** program. It can take huge data sets and perform analyses that are considered in "data mining" field. – ttnphns Jan 08 '12 at 13:13
  • @ttnphns why does that matter? Data mining is also statistics. Note I also said "seems like a great tool for data mining.." – Tom Jan 08 '12 at 17:01
  • 8
    I love the comment "it allows you to do advanced computations without having the... knowledge required...". This is, in fact, the primary competitive advantage of SPSS products, and reflects its market strategy; in exchange for this you pay a hefty price, as you note. However, being able to run advanced analyses without the required understanding does not bode well for the quality of the resulting findings. Although R requires a steep, up-front learning curve, it will benefit you in the long run. In addition, there are plenty of tutorials and learning materials available, and it's free. – gung - Reinstate Monica Jan 08 '12 at 19:33
  • 1
    @gung, that does seem pretty obvious though. Nevertheless, this software may be great for quickly discovering all kinds of relationships between data. I imagine it taking a lot more time in R to prepare such tests, which would be a disappointment if no relationships are found. – Tom Jan 09 '12 at 13:43
  • 3
    That's true. SPSS is *much* more user friendly than R; you can run many analyses quickly and easily with its point and click interface. That can be nice sometimes, but is it worth the money? I think in the long run it's worth your time to learn R. – gung - Reinstate Monica Jan 09 '12 at 15:12
  • 3
    I agree with gung. As a manager, I would rather spend the 20,000 euros on training staff in R and a better understanding of statistics. Any glib, menu-driven data mining is highly prone to giving misleading results. – Peter Ellis Jan 10 '12 at 19:57
  • 2
    From a recent Rexer survey: "IBM SPSS Modeler (SPSS Clementine), Statistica, and IBM SPSS Statistics (SPSS Statistics) are identified as the primary tools used by the most data miners. Open source tools Weka and R and are increasingly used by large numbers of both academic and for-profit data miners. Users of IBM SPSS Modeler, Statistica, and Rapid Miner are the most satisfied with their software. SAS Enterprise Miner dropped in data miner’s tool rankings this year. Information on obtaining a free copy of the report is available at RexerAnalytics.com." – rolando2 Jan 12 '12 at 01:50
  • @rolando2 if so many people use SPSS Modeler, why are there no answers to this question yet? – Tom Jan 12 '12 at 10:26
  • Gung, its not the speed in terms of power thats important in most commercial situations. Its the speed of the analytics process that's important. The speed at which you run through various sets of analysis, the speed at which you can back track and try alternative approaches, the speed at which you can action those changes. This is where modeller beats not only R but all the other packages as well because this speed of analysis is at the core of its design and interface. But yes. R is cheaper. Why not spend a year or two learning that and then take 5 times longer to actually do anything useful –  Sep 28 '12 at 15:26
  • I appreciate your perspective, @Mickey, & you make some valid points here. On a different note, this isn't really an answer to the question the OP posed. You may want to read our [FAQ](http://stats.stackexchange.com/faq). – gung - Reinstate Monica Sep 28 '12 at 16:53

2 Answers2

5

My experiences with SPSS Modeler are very positive. It allows you to build complex systems by simply connecting and disconnecting nodes, and it has a nice graphical representation of the node network that you have built. I prefer seeing the whole "picture", as it is easy to get lost in code if you were to code everything yourself and if things go very complex.

There are nodes for manipulating with rows, columns, outputs and inputs, deriving new features, various modelling nodes, charting and much more. To me it serves as a multi-purpose tool, because I can use SQL to query directly from a database (through an ODBC connection) or use text files (and transform them if broken) if a database connection is not present. It even supports Excel files.

It has a built-in legacy scripting language; a formula builder (to build your logic for calculations, node-specific manipulations, selections etc.); and since version 16 it also supports scripting in Python (scripting in SPSS Modeler is mostly used for automation of tasks - say, you need to regularly perform some calculations as part of some ETL process).

As for statistical modelling, there are a bunch of modelling specific nodes (categorized under clusters, classifiers, networks) which can be fine-tuned with parameters (just as you would define your parameters in R, Python or elsewhere). You can even use R in a modelling node (a specific node for R). Those models can be saved, re-trained, compared etc. There are even auto-classifiers or auto-clusters that try a bunch of stuff and then let you choose the best model (which mostly, of course, varies on input data).

I believe this software is incomparable to other GUI based statistical modelling software and justifies its price.

-- 3 years late to reply, but maybe this helps someone :)

Nick Cox
  • 48,377
  • 8
  • 110
  • 156
kasparg
  • 101
  • 1
  • 5
0

try this website http://www.ibm-spss-modeler.com?EZE I found some usefull information about IBM SPSS Modeler there. You can also start a free trial of Modeler directly from the site. Hope this will help.

user143676
  • 1