Hybrid Recommendation System Memanfaatkan Penggalian Frequent Itemset dan Perbandingan Keyword
Wayan Gede Suka Parwita(1*), Edi Winarko(2)
(1) 
(2) 
(*) Corresponding Author
Abstract
Abstrak
Recommendation system sering dibangun dengan memanfaatkan data peringkat item dan data identitas pengguna. Data peringkat item merupakan data yang langka pada sistem yang baru dibangun. Sedangkan, pemberian data identitas pada recommendation system dapat menimbulkan kekhawatiran penyalahgunaan data identitas.
Hybrid recommendation system memanfaatkan algoritma penggalian frequent itemset dan perbandingan keyword dapat memberikan daftar rekomendasi tanpa menggunakan data identitas pengguna dan data peringkat item. Penggalian frequent itemset dilakukan menggunakan algoritma FP-Growth. Sedangkan perbandingan keyword dilakukan dengan menghitung similaritas antara dokumen dengan pendekatan cosine similarity.
Hybrid recommendation system memanfaatkan kombinasi penggalian frequent itemset dan perbandingan keyword dapat menghasilkan rekomendasi tanpa menggunakan identitas pengguna dan data peringkat dengan penggunaan ambang batas berupa minimum similarity, minimum support, dan jumlah rekomendasi. Nilai pengujian yaitu precision, recall, F-measure, dan MAP dipengaruhi oleh besarnya nilai ambang batas yang ditetapkan.
Kata kunci— Hybrid recommendation system, frequent itemset, cosine similarity.
Abstract
Recommendation system was commonly built by manipulating item is ranking data and user is identity data. Item ranking data were rarely available on newly constructed system. Whereas, giving identity data to the recommendation system causes concerns about identity data misuse.
Hybrid recommendation system used frequent itemset mining algorithm and keyword comparison, it can provide recommendations without identity data and item ranking data. Frequent itemset mining was done using FP-Gwowth algorithm and keyword comparison with calculating document similarity value using cosine similarity approach.
Hybrid recommendation system with a combination of frequent itemset mining and keywords comparison can give recommendations without using user identity and rating data. Hybrid recommendation system using 3 thresholds ie minimum similarity, minimum support, and number of recommendations. With the testing data used, precision, recall, F-measure, and MAP testing value are influenced by the threshold value.
Keywords— Hybrid recommendation system, frequent itemset, cosine similarity.
Full Text:
PDFDOI: https://doi.org/10.22146/ijccs.7545
Article Metrics
Abstract views : 3674 | views : 3555Refbacks
- There are currently no refbacks.
Copyright (c) 2015 IJCCS - Indonesian Journal of Computing and Cybernetics Systems
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
View My Stats1