ABSTRAKSI: Perkembangan internet yang sangat pesat menyebabkan terjadinya information overload, dimana user mendapat kesulitan jika ingin mendapatkan sesuatu yang benar-benar diperlukan.. Salah satu solusi untuk mempermudah user mencari informasi yang dibutuhkan adalah Recommender System. Recommender system adalah sebuah sistem yang dapat memberikan rekomendasi berupa prediksi rating terhadap suatu item berdasarkan persamaan karakteristik user dalam memberikan informasi.
Oleh karena itu, tugas akhir ini mengimplementasikan dan menganalisis user-based collaborative filtering recommender system, yang menerapkan Pearson Correlation dan Spearman Correlation. Pearson Correlation dan Spearman Correlation digunakan untuk mengolah nilai rating user. Tujuan dari tugas akhir ini adalah untuk menganalisis akurasi prediksi rating yang dihasilkan oleh recommender system setelah diimplementasikan kedua korelasi. Parameter yang digunakan dalam analisis adalah ukuran threshold co-rated items, perbandingan training set dan test set serta ukuran neighbourhood yang diukur dengan Mean Absolute Error.
Hasil pengujian menunjukkan bahwa akurasi prediksi yang dihasilkan oleh Pearson Correlation dan Spearman Correlation semakin meningkat dengan bertambahnya jumlah co-rated items. Semakin besar ukuran neighbourhood, akurasi prediksi yang dihasilkan juga semakin baik. Nilai error dari prediksi semakin menurun dengan bertambahnya jumlah training set . Hasil prediksi kedua korelasi menunjukkan bahwa nilai error yang dihasilkan Pearson Correlation lebih rendah dibandingkan Spearman Correlation.
Kata Kunci : recommender system, collaborative filtering, Pearson Correlation, Spearman CorrelationABSTRACT: The internet development these days with too much information cause information overload, where users have a difficulty in making decisions by the presence of too much information. One of the solution to ease the users finding the required information is recommender system. Recommender system is a system that can give recommendation in term of rating prediction of an item, based on the similarity of the user characteristic in giving information.
Therefore, in this final project, the author implement and analyze user-based Collaborative Filtering Recommender System, which applies Pearson Correlation and Spearman Correlation. Pearson Correlation and Spearman Correlation is used to compute user’s rating in giving similarity. The purpose of the final project is to analyze prediction accuracy result that is given by the recommender system after both of the correlation is implemented. The parameter that is used in the analysis is the threshold co-rated items size, the neighbourhood size, and the comparison between training set and test set, measured by Mean Absolute Error.
Based on the analysis, Pearson Correlation and Spearman Correlation produces prediction accuracy that increases with the increase number of co-rated items. The bigger the neighbourhood size, the better the prediction accuracy is given. The prediction accuracy also get better with the bigger of the training set size. The recommendation of both correlation shows that the differences value of error rate is relatively small, with Pearson Correlation’s error rate smaller than Spearman Correlation.Keyword: recommender system, collaborative filtering, Pearson Correlation, Spearman Correlation