Skill set of a data miner

[minero]

Hi, I am currently an Oracle DBA. I've read some stuff about data mining and it seems to be an interesting career. I plan on pursuing such career, like taking MS in Statistics, only if my being an Oracle DBA is a requirement or a great asset skill to be a data miner.

If anyone could be so kind enough to clarify the following regarding the skill set of a data miner:

Is experience in Database Administration a required skill?
Is just being a Statistic graduate (no experience as a DBA) enough to be a data miner?
What are the usually required IT/software skills to be a data miner?
What is the typical skill set of a good data miner?

I hope someone replies to this query. It would be alot of help in my career planning. Thank you very much.

Regards,



Dyle

[editor]

Your experience with databases and as DBA is helpful, but being DBA is not required for data mining.
Knowledge of statistics (whether with Stat degree or not) is definitely required.

Being a good data miner is more of a mindset than skill set.
For example, if you enjoy solving puzzles, are naturally curious,
and enjoy learning new things than you will probably be a good data miner.

There is a proposed data mining curriculum developed
by SIGKDD - which describes the relevant topics see
http://kdd.org/curriculum.php

There are also several places that offer M.S. in data mining - see
www.kdnuggets.com/education/

[TimManns]

I would agree that being a DBA is not a necessity, bit it will certainly help. I use SPSS Clementine and Teradata, and analyse mobile phone usage data for things like, churn/attrition, cross-sell, up-sell etc. Most of my work runs as auto-generated sql within Teradata.

I would say that understanding basic sampling method is very important. I don't have a statistical background, and made some samplying errors in my first projects. I think you can pick up and statistics components of data mining fairly quickly.

Because I am dealing with many millions of rows of data, for me it is *very* important that I use indexes, summarise data correctly and join multiple database tables using the correct keys and attributes. Otherwise the queries will be ineffecient and might take hours. This does require some basic understanding of the database.

In my view a good dataminer is able to translate the business requirements (churn, cross-sell etc) into the required data analysis. This would be the most important aspect.
Secondly, indentify the data sources required for the project, ensure the database has capacity and is organised correctly to do this (if the DB wasn't designed for analysis it might even involve changing indexes for example, or submitting a requirement for additional data to be collected or stored in the DB).
Thirdly, conduct the analysis in a timely manner and communicate with your collegues (this might be a trade off between elapsed time and success of results, sometimes i settle for lesser results if my project can complete in half the time).
Lastly, presentation skills!. You have to get the executives and business sponsors to understand and *like* the work you have done. Prove the business case for the project. Avoid lots of numbers and statistics (this is where many hard-core statisticains fail), and keep to simple graphs and % of customers or money saved as a result of the project.

Hope this helps

Tim