A Supervised learning algorithm which can classify multi-class problems. Brief Description: A random forest classifcation algorithm which used Kmeans ro perform data splits at each node ( the splits are unsupervised , and donot use labels)
-
Data format: H (n*m) input matrix with n points and m features .
-
Labels (n*1) - Labels to corresponding data files.
-
Creating Train and Test datasets for multiple runs. You can use the "supervised_bagging" function of this purpose. (Let me know if any difficulty)
-
TrainDMapIndices - Indices of randomly shuffled training points.
-
TrainLabels - Corresponding training labels.
-
TestDMapIndices - Indices of randomly shuffled testing points.
-
TestLabels - Corresponding testing labels.
-
Other parameters in the file can be changed according to the need.
Please let me know before any major changes.