"Neural Net - Lot of questions"

rajanm

Hello All,

The problem is to predict a sales trend based on the data available. I am looking at using the supervised neural net for solving a problem. I am using rapidminer 4.4 community edition.

Attached are the files I am using.

- input data containing store, day, sales data, customers etc and the aml file
- the parameters xml file.
- time series showing the classification errors
- classification error details

My questions are (some of them related to the neural net)
- are the right pre processing (discretization, normalization) being done on the data?
- are the right parameters used in the neural net?
- how does one know if the neural net has been trained adequately?
- are there any rapid miner options which can train the neural net faster?
- are there any options to gauge how accurate the model is?

i tried applying this neural net model on another input data and the accuracy was only 50% 
I think it's something to do with the attributes in the data.

Thanks.


[attachment deleted by admin]