PERBANDINGAN EFEKTIFITAS METODE USER-BASED COLLABORATIVE FILTERINGDENGAN METODE USER-ITEM BASED COLLABORATIVE FILTERING

AGUSTA, INDIKA SATRIYANA (2013) PERBANDINGAN EFEKTIFITAS METODE USER-BASED COLLABORATIVE FILTERINGDENGAN METODE USER-ITEM BASED COLLABORATIVE FILTERING. Other thesis, Universitas Sebelas Maret.

[img]
Preview
PDF ( testing set, user-based collaborative filtering, user-item based collaborative filtering) - Published Version
Download (796Kb) | Preview

    Abstract

    This research will compare the effectiveness user-based collaborative filtering method and user-item based collaborative filtering method. User-based collaborative filtering method provides high-quality recommendation because it involves user subjectivity, but its weakness lies in sparsity and scalability. Useritem based collaborative filtering method can provide recommendation without looking for neighborhood formation. However, if many predicted value user-item based collaborative filtering is beyond rating value interval, so the accuracy level becomes weak. The comparison was carried out on the sample data consisting of 10000 ratings constituting the Jester’s continuous ratingfrom -10 to 10 containing 100 users and 100 items. The examination was repeated 30 times for each testing set level to obtain the average of NMAE and predicted time. The user-based collaborative filtering method was divided into two based on the number of neighborhood. User-based collaborative filtering containing 10 neighborhoods (N-10) and user-based collaborative filtering had similarity threshold > 0.3. The result of examination it could be found that the average values of NMAE and predicted time of user-based collaborativefiltering method with N-10, user-based collaborative filtering with similarity threshold > 0.3, and user-item based collaborative filtering were 0.1850; 59 s, 0.1854; 111 s, 0.1870; 29 s. From the viewpoint of NMAE, user-based collaborative filteringwith N-10 was more effective, however from the viewpoint of predicted time, user-item based collaborative filtering was more effective. Keywords: testing set, user-based collaborative filtering, user-item based collaborative filtering

    Item Type: Thesis (Other)
    Subjects: Q Science > QA Mathematics > QA76 Computer software
    Divisions: Fakultas Matematika dan Ilmu Pengetahuan Alam > Informatika
    Depositing User: Lia Primadani
    Date Deposited: 10 May 2014 18:36
    Last Modified: 10 May 2014 18:36
    URI: https://eprints.uns.ac.id/id/eprint/15945

    Actions (login required)

    View Item