Due to recent updates, all users are required to create an Altair One account to login to the RapidMiner community. Click the Register button to create your account using the same email that you have previously used to login to the RapidMiner community. This will ensure that any previously created content will be synced to your Altair One account. Once you login, you will be asked to provide a username that identifies you to other Community users. Email us at Community with questions.
"svm output range very differnt from neural net"
Learner III
Hi,
Firstly let me thank you for a wonderful product, ive being playing with for a few days and its a lot of fun, if a bit overwhelming to start.
As a hobby Ive been using nueral nets to predict horseracing results for a while now and wanted to give svm a try.
the svm is doing well but the output range is very strange. My target is win (1) or lose(0), so i expect to get back some number between 0 and 1.instead I get ranges between -2 and 10 and the value is inverted, so for a winner it returns ~-2 and for a loser ~10.
Any idea how I can get the svm to scale the output like a nn?
Firstly let me thank you for a wonderful product, ive being playing with for a few days and its a lot of fun, if a bit overwhelming to start.
As a hobby Ive been using nueral nets to predict horseracing results for a while now and wanted to give svm a try.
the svm is doing well but the output range is very strange. My target is win (1) or lose(0), so i expect to get back some number between 0 and 1.instead I get ranges between -2 and 10 and the value is inverted, so for a winner it returns ~-2 and for a loser ~10.
Any idea how I can get the svm to scale the output like a nn?
Tagged:
0
Answers
actually it seems to me that you are trying to solve a classification problem (win or loose) by a regression. RapidMiner does all the handling of ranges and probabilities internally if you just set the label to a nominal value. So if you use the numerical to binominal operator to change the data accordingly, SVM as well as NeuralNet will produce a crisp classification and a confidence for each class.
Greetings,
Sebastian