how to use Semi-supervised Learning algorithms in Rapidminer?

mkq

hi,all.

I want to use Semi-supervised Learning algorithms in Rapidminer,such as self-training、generative models、SVMs、graph-based methods、multiview learing. But I can't find them in Rapidminer.How can I find them?

Best Wishes,
mkq

MartinLiebig

Hi,

The SVM has an option to go one-class. That might help you.

Best,
Martin

mkq

Hi, Martin!

Thank you very much! But one-class SVM seems to be unsupervised learning algorithm! I want to use Semi-supervised Learning algorithms.

Best Wishes,
mkq

MartinLiebig

have you checked the weka extension? To be honest i never used those algorithms.

If there is a java implementation we (= community?) might embedd it. It's pretty straight forward.

~Martin

JEdward

Here's a couple of papers I tracked down on Google Scholar which might give you some places to start.

http://link.springer.com/article/10.1631/jzus.C1000330 -  Clustering feature decision trees for semi-supervised
classification from high-speed data streams - Wen-hua XU, Zheng QIN, Yang CHANG
SmSCluster is a semi-supervised method they implemented using RapidMiner.

http://arxiv.org/ftp/arxiv/papers/1201/1201.1670.pdf - Customers Behavior Modeling by Semi-Supervised Learning inCustomer Relationship Management - Siavash Emtiyaz, MohammadReza Keyvanpour
"The proposed semi-supervised method is a model by means of a feed-forward neural network trained by a back propagation algorithm (multi-layer perceptron) in order to predict the category of an unknown customer (potential customers). In addition, this technique can be used with Rapid Miner tools for both labeled and unlabeled data"
(1-class SVM should work with this approach too)
Distance measures, clusters, HBOS + stacking can all be used too. Howeyou might need to drop to R, Python or a Java library to implement others.

Regarding graph models, the Linked Open Data extension is probably a good place to start. 

mkq

Hi, Martin!

Thank you very much! I'll  check the  java implementation and weka extension.

Best Wishes,
mkq

mkq

Hi, JEdward!

Thank you very much! I'll try the methods you've advised.  ;D ;D


Best Wishes,
mkq