• Login
    View Item 
    •   USU-IR Home
    • Faculty of Computer Science and Information Technology
    • Department of Computer Science
    • Doctoral Dissertations
    • View Item
    •   USU-IR Home
    • Faculty of Computer Science and Information Technology
    • Department of Computer Science
    • Doctoral Dissertations
    • View Item
    JavaScript is disabled for your browser. Some features of this site may not work without it.

    Hybrid Approach Redefinition (HAR) Method dalam Permasalahan Class Imbalance

    View/Open
    fulltext (6.586Mb)
    Date
    2018
    Author
    Hartono
    Advisor(s)
    Sitompul, Opim Salim
    Tulus
    Nababan, Erna Budhiarti
    Metadata
    Show full item record
    Abstract
    In the classification, the dataset is said to be imbalanced when there is a class with a smaller amount of data than the other classes. Class imbalance is a problem because machine learning will produce a good classification accuracy of the class with a large number of members (majority class), while the class with the number of minor members (minority class) has a poor accuracy when minority class often become an interesting class to be observed (positive class). One of the existing methods is the Hybrid Ensembles method which is an ensemble learning method that combines bagging and boosting. In the handling of class imbalance it is necessary to pay attention to the number of classifier and diversity (diversity) data. The results show that Hybrid Ensembles require a large number of classifier and certainly need to maintain data diversity. This research is intended to optimize the results of the hybrid ensemble method, especially in reducing the number of classifier and also maintaining good data diversity. The number of classifier will be reduced through the application of preprocessing using Random Balance Ensemble Method method and also the data diversity can be maintained by using Different Contribution Sampling method. The results show that Hybrid Approach Redefinition (HAR) Method can reduce the number of classifier and maintain data diversity.
     
    Di dalam klasifikasi, dataset dikatakan tidak seimbang (imbalanced) ketika terdapat suatu class dengan jumlah data yang yang lebih kecil dibandingkan dengan class yang lain. Class imbalance merupakan masalah karena machine learning akan menghasilkan akurasi klasifikasi yang baik terhadap class dengan jumlah anggota yang banyak (majority class), sedangkan kelas dengan jumlah anggota sedikit (minority class) memiliki akurasi yang kurang baik padahal minority class sering menjadi class yang menarik untuk diamati (positive class). Salah satu metode yang ada adalah metode Hybrid Ensembles yang merupakan metode ensemble learning yang menggabungkan bagging dan boosting. Di dalam penanganan class imbalance tersebut perlu memperhatikan jumlah classifier dan keanekaragaman (diversity) data. Hasil penelitian menunjukkan bahwa Hybrid Ensembles memerlukan jumlah classifier yang besar dan tentunya perlu mempertahankan keanekaragaman data. Penelitian ini dimaksudkan untuk melakukan optimisasi terhadap hasil dari metode hybrid ensemble khususnya di dalam mengurangi jumlah classifier dan juga mempertahankan keanekaragaman data yang baik. Jumlah classifier akan dikurangi melalui penerapan preprocessing dengan menggunakan metode Random Balance Ensemble Method dan juga keanekaragaman data dapat dipertahankan dengan menggunakan metode Different Contribution Sampling. Hasil penelitian menunjukkan bahwa Hybrid Approach Redefinition (HAR) Method dapat mengurangi jumlah classifier dan mempertahankan keanekaragaman data.

    URI
    http://repositori.usu.ac.id/handle/123456789/4938
    Collections
    • Doctoral Dissertations [51]

    Repositori Institusi Universitas Sumatera Utara (RI-USU)
    Universitas Sumatera Utara | Perpustakaan | Resource Guide | Katalog Perpustakaan
    DSpace software copyright © 2002-2016  DuraSpace
    Contact Us | Send Feedback
    Theme by 
    Atmire NV
     

     

    Browse

    All of USU-IRCommunities & CollectionsBy Issue DateTitlesAuthorsAdvisorsKeywordsTypesBy Submit DateThis CollectionBy Issue DateTitlesAuthorsAdvisorsKeywordsTypesBy Submit Date

    My Account

    LoginRegister

    Repositori Institusi Universitas Sumatera Utara (RI-USU)
    Universitas Sumatera Utara | Perpustakaan | Resource Guide | Katalog Perpustakaan
    DSpace software copyright © 2002-2016  DuraSpace
    Contact Us | Send Feedback
    Theme by 
    Atmire NV