ABSTRAKSI: Recommender system merupakan sebuah sistem yang dapat digunakan untuk memprediksi sebuah item berdasarkan informasi yang diperoleh dari user, sehingga didapatkan rekomendasi berdasarkan profil penggunanya
Tugas Akhir ini mengimplementasikan dan menganalisis Content-Boosted Collaborative Filtering (CBCF) yang merupakan recommender system yang menggabungkan antara content based filtering dengan collobarative filtering. Penggabungan pendekatan content based filtering dengan collaborative filtering pada metode CBCF bertujuan untuk menanggulangi kekurangan yang ada pada kedua pendekatan sebelumnya terutama First-Rater Problem dan Sparsity. Tugas akhir ini menganalisis akurasi prediksi rating yang dihasilkan oleh recommender system setelah diimplementasikan metode CBCF. Parameter yang digunakan dalam analisis adalah sparsity rate dan jumlah neighborhood.
Akurasi prediksi yang dihasilkan oleh metode CBCF lebih besar dibandingkan dengan pure collaborative filtering. Performansi terbaik terjadi saat jumlah neighborhood mendekati jumlah training user dan tidak terdapat missing rate. Hasil rekomendasi pada metode CBCF pada recommender system menunjukkan kesesuaian antara genre item hasil rekomendasi dengan genre item yang telah diberi rating oleh active user.Kata Kunci : recommender system, collaborative filtering, metode CBCF, sparsity, first-rater problem, genreABSTRACT: Recommender system is a system that can be used to predict the items based on information obtained from users, so we get recommendations based on user profiles
In this final project, the implemention and the analyzing Content-Boosted Collaborative Filtering (CBCF), which is a recommender system that combines the approaches between content-based filtering with collobarative filtering. The purpose of combination between content based filtering and collaborative filtering in CBCF for overcome the existing shortcomings in the two previous approaches, especially the First-Rater Problem and Sparsity. This final rating analyze prediction accuracy generated by the recommender system was implemented CBCF method. The parameters used in the analysis is the sparsity rate and the number of neighborhood.
Prediction accuracy generated by CBCF method is better than pure collaborative filtering. Best performance occurs when the number of neighborhood near the amount of training user and there are no missing rate. Result of the recommendations on the method of CBCF method in recommender system shows the compatibility between the item genre from result of recommendation with item genre that has been rated by active user.Keyword: nder system, collaborative filtering, CBCF method, sparsity, first-rater problem, genre