View Full Version : Predictive Modeling using WEKA

05-31-2009, 11:37 PM

I have some knowledge to Data miner tool- "WEKA". I just wanted to get a brief explanation about "Implementing Predictive modeling using WEKA".:confused:

I also wanted to know the difference betweeen the "Output of the Cross validation stratergy and Output of the training set", how these two are related to each other.......?

I request anyone who has some knowledge about this to please help me to resolve my problem....... I will be thankful to you guys if you please exaplain me the above queries:confused:.

Thankin you :D

06-02-2009, 07:19 PM
Hi Naveen,

You might need to be a bit more specific with your first question. Do you want to find out about how to develop a predictive model using Weka? Or how to deploy a predictive model once you've learned one that meets requirements? Or perhaps, how to write a new learning algorithm in the Weka framework?

Cross-validation is a statistical technique that estimates the performance of a learning scheme on fresh data when you have limited training data. Normally, if you have plenty of data, you'd set aside some of it in a hold out set to evaluate the performance of your learning algorithm. When you have limited data, cross-validation can make the most of your data by dividing it up into k distinct chunks (folds) and using each fold in turn to evaluate the model that has been learned using the remaining k-1 folds. The performance over the k test folds is averaged to give an estimate of how well the scheme will perform on new (as yet unseen) data.

Looking at performance on the training set is not useful for getting a feeling for how the model will perform in the future on new data. This is because the training data has been used to learn the model and this estimate will be overly optimistic. It is useful, when compared to the performance under cross-validation or a hold out set, for getting a feeling for whether the learning algorithm is over fitting the data.

Hope this helps.


06-04-2009, 10:48 PM
Hello sir:),

Thanks for your reply. I understood the difference. Actually i wanted to know "HOW THE PREDICTIVE MODEL IS DEVELOPED IN WEKA AND HOW IS IT ANALYSED ?":confused: I will also like to know the remaining two questitions put by you, but at present i am interested inknowing about predictive model.:D

I will be pleased if you explain me about the developing procedure of the predictive model ?

Thanking you:)