Can any one tell me how I choose a data mining technique
Does anyone have examples of when they think they have applied the wrong
tecnique

There's not really a pat answer for this as it really depends on what you
are looking for.

Do you need to do predictions or are you just looking for data description?
Are you looking for the best prediction accuracy? Are you looking for the
best lift? Do you need descriptive rules to explain the predictions? What
training and query performance do you require?

All these questions help determine which technique is best for you. For
example, Naive Bayes, Decision Trees, and Neural Nets all are typically
classification algorithms. Naive Bayes is the fastest for training, while
Decision Trees is the fastest for prediction - particularly if you have many
inputs. Neural Nets is slower for both, but can find patterns that will be
missed by the other two. The results if Naive Bayes and Decision Trees are
fairly easy to understand, while Neural Nets are fairly impossible.

If accuracy is the most important issue, you can build several models and
use a lift chart/classification matrix to determine which performs the best.
These charts are available in SQL Server 2005. For SQL Server 2000 you can
download the Data Mining resource kit which has a lift chart implemented by
Angoss.

--

-Jamie MacLennan
SQL Server Data Mining
This posting is provided "AS IS" with no warranties, and confers no rights.
[quoted text, click to view]