ABSTRAKSI: Recommender system adalah sebuah aplikasi yang dapat memberikan rekomendasi berupa prediksi rating terhadap suatu item berdasarkan persamaan karakteristik user dalam memberikan informasi.
Tugas akhir ini mengimplementasikan dan menganalisis item-based collaborative filtering recommender system, yang menerapkan algoritma slope one dan weighted slope one. Tugas akhir ini menganalisis akurasi prediksi rating yang dihasilkan oleh recommender system setelah diimplementasikan algoritma slope one dan weighted slope one. Parameter yang digunakan dalam analisis adalah perbandingan training set dengan test set. Selain itu, tugas akhir ini juga menganalisis kesesuaian hasil rekomendasi dengan tipe atau konten dari item yang direkomendasikan.
Akurasi prediksi yang dihasilkan oleh algoritma slope one dan weighted slope one akan semakin meningkat dengan bertambahnya jumlah data yang terdapat pada training set. Hasil rekomendasi pada algoritma slope one dan weighted slope one menunjukkan ketidaksesuaian antara tipe item hasil rekomendasi dengan tipe item yang telah diberi rating oleh active user. Hal ini disebabkan proses prediksi rating pada algoritma slope one dan weighted slope one tidak memperhatikan tipe atau konten dari item, tetapi lebih memperhatikan kemiripan pola rating.
Kata Kunci : Kata kunci: recommender system, collaborative filtering, algoritma slope oneABSTRACT: Recommender system is an application that can give a recommendation in term of rating prediction of an item, based on the similarity of user characteristic in giving information.
In this final project, the implementation and the analysis of the item-based collaborative filtering recommender system, which applies slope one and weighted slope one algorithm, is performed. Initially, slope one and weighted slope one algorithm is implemented into the recommender system. Then, the analysis is carried out to the accuracy of rating prediction result that is given by the recommender system. Comparison parameter on training set and set test is used in the analysis. Furthermore, in this final project, suitability analyses between recommendation result and the recommended type or the recommended content is performed.
The accuracy of prediction, which is resulted by both algorithm, increases with the increase of the number of data in the training set. In the result of recommendation analyses, it is founded unsuitability between item types of recommendation result and item types that have been rated by an active user. This condition probably is caused by prediction process of the rating in both algorithm that prefer the similarity of rating pattern to types or contents of an item.
Keyword: Keywords: recommender system, collaborative filtering, slope one algorithm