MRMD: Maximum-Relevance-Maximum-Distance



MRMD

Feature selection method has two main part of the decision: The Pearson correlation coefficient shows the closely relationship between features and labels.Distance between features is used to measure the redundancy. Finally, MRMD selects features which has strong correlation with labeled and lowest redundancy features subset Pearson

Usage

mrmd.jar( Download the jar of MRMD )


java -jar mrmd.jar -i input -o output.txt


  • -i inputFile
  • -o outputFile
  • Parameters
    (1)-df:
    the distance function default(1)
    (1 = Euclidean Distance, 2 = Cosine Distance, 3 = Tanimoto coefficient, 4 = mean)
    (2)-a:
    outputfile of arff default (opt.arff)
    (3)-m:
    opt model type default(rf)
    (rf, svm, bagging)
    (4)-sn 100
    (5)-N (no opt)
    (6)-t: Turn on multithreading(default 1),You can use the linux command to view the threads on the machine(grep 'processor' /proc/cpuinfo | sort -u | wc -l)
    (7)-c the csv file name(it includes accuracy and f1_score data)


    Download

    We have got our project on Github(https://github.com/heshida01/mrmd_mulan.git)

    All Rights Reserved Copyright @ 2016|Prof. Quan Zou
    Last Modified in 2018/7/17