Dear forum users,
I am a newbie in using MS SQL server with analysis services.
There seems to be no 'cross-validation' tool in MS SQL
which is frequently used in data mining and even statistics.
Is there anyone having similar difficulties?
Is there any solution like a small scripts to divide
the given dataset with multiple folds?
Your valuable comments and feedbacks would be appreciated.
Minnetongka
We have implemented cross-validation as part of a larger "model evaluation" module on top of the data mining algorithm implementations in SQL-Server 2005 Analysis Services. This is done by implementing sampling on top of the source case/nested tables and then utilizing the Analysis Services APIs (in C#) to train predictive models over training sets and then execute the appropriate prediction join over the testing sets to collect model performance metrics.
We commonly use this "model evaluation" model to automate the process of finding the most accurate predictive models for a given application.
Although we can't make our module publicly available at this time, let me know if you're interested in more information.
Thanks,
- Paul Bradley
No comments:
Post a Comment