|
[x]
Posted via EE Mobile
|
|
| Search, ask, and monitor your questions on the go with EE Mobile. Visit Experts Exchange from your mobile device and never be out of touch again. |
|
|
|
|
Asked by KlausRoeder in Data Mining, Probability & Statistics, Math & Science
I plan to use a neural net for prediction.
My goal is a classification (3 classes) providing the propability of the outcome of each class.
There will be about 3000 to 10000 training cases with approx. 5 to 15 predictors.
The main focus is on the propability. There is no need to assign a case to a class clearly or to calculate assignment errors.
What form of neural nets (backpropagation etc. is suitable in this usecase? Maybe you do not only give me the neural net form, but also the appropriate programming tool (Java preferred - e.g. JOONE). Possibly the neural net form generally supports specifying the propability of the classification but the tool doesn't show it.
20091111-EE-VQP-92 - Hierarchy / EE_QW_2_20070628