Tools for Data Mining

[plank123]

Hi,

Currently I am trying to evaluate and compare the various tools available for data-mining. However, I want the selected software's to be from the 'statistical' point of view. Thus, I dont want to review the general softwares for data-mining, but rather focus on Statistical-Tools for data mining.

:?: I am focusing on SAS, SPSS, Statistica, etc. Am I on the right track? Could you please mention other REPUTED companies in this segment? I found many companies on the kdnuggets site, but want to pick out 5 from them. Can I include IBM, Oracle, or Microsoft in this segment?

Thanx a ton for your advice! :)

[TimManns]

Hi,

I would suggest that your main consideration be your existing infrastructure. What systems do you currently have within your company, and how well will any new data mining solution integrate with this?

I used to work at SPSS, now working in a telco doing churn analysis and complex ad-hoc analysis. We use Clementine, with all our data stored on a Teradata system. We don't have an intermediate level of summary data, we access our data from fairly detailed database tables and might create summary tables as part of our analysis itself. Very rarely is our analysis the same twice (mostly because the market changes so quickly). Some data mining solutions might require a data mart be created in a proprietary format, or a duplication of data be stored. This can take effort. For some situations having a data mart or off-line copy of data is a bonus.

Also consider your staff skillsets. Are they statisticians, DBA's, data analysts, or data miners (I believe there are subtlie differences). Training time and adoption of a new tool will take time, so using something easy or familar is a bonus. Most of our skills are in SQL and data analysis.

On my home PC I've played around with SQL Server 2005. I was able to build a good data mining solution but it took more effort and programmer skills. If you are developing a single solution that will be pretty static, then I think this investment is fine. In this situation a database related solution from Oracle, SQL server etc is probably ideal. If you intend on having fast turn around of many data mining projects, then you'll probably safer opting for a tool from SPSS, SAS, or STATSOFT.

Hope this helps

Tim