LibD3C

LibD3C: Ensemble classifiers with a clustering and dynamic selection strategy

LibD3C

LibD3C employs two types of selective ensemble techniques, which are a combination of the ensemble pruning based on k-means clustering and dynamic selection and circulating combination.We upgrade the software to LibD3C to handle imbalance data problem.The input file should be a .arff file. You can use this tool in any OS with JVM. software can only identify .arff file

Usage

downLoad usage.ppt

doc_usage.zip( Download the jar of libD3c and example )

Video

Please use the IE browser to play

CrossValidation

Balance data

Java -jar LibD3C.jar -c fold-num trainFile

Imbalance data

Java -jar LibD3C.jar -m -c fold-num trainFile

-c means crossvalidation fold-num is the crossValiditon folds. trainFile is the file path for the training data

prediction

(1)train model

Balance data

Java -jar LibD3C.jar -t trainFile

Imbalance data

Java -jar LibD3c.jar -m -t trainFile

-t means train model and the model would be saved in the same filePath(a new file "train.model" will be created) ,trainFile is the file path for the training data,resultFile is the file path save prediction result

(2)predict instance

Balance data

Java -jar LibD3C.jar -p train.model testFile resultFile

Imbalance data

Java -jar LibD3C.jar -m -p train.model testFile resultFile

-p means prediction and train.model is the filePath you trained before ,trainFile is the file path for the training data, resultFile is the file path save prediction result

Download

You should download LibD3C.jar and classifier.xml.You should also put them together(in the same file path).We have got our project on Github:

Version	Cluster Algorithm	Download Address
LibD3C1.1	KMeans	https://github.com/smilida/LibD3C_1.1
LibD3C1.2	KMeans	https://github.com/smilida/LibD3C_1.2
LibD3C2.0	Affinity Propagation	https://github.com/smilida/LibD3C_2.0