ABSTRAKSI: Recommender system dengan metode Collaborative Filtering terkadang tidak akurat dikarenakan data sparsity. Salah satu metode untuk menambah ketepatan akurasi dalam perhitungannya adalah menggunakan Metode Prediction Error-Based Enhancement with Counting Number of Common Neighbors (PEBE-CN).Metode ini merupakan pengembangan dari user-based collaborative filtering yaitu User-Based Pearson Similarity (UBPS). Metode ini memprediksi dengan melihat prediction error dari item-item yang telah di-rate oleh active user, dan diberi bobot berdasarkan common neighbornya pada desired item. Beberapa hal yang diteliti antara lain pengaruh parameter n, γ, dan perbandingan training set dengan test set terhadap akurasi prediksi menggunakan mean absolute error (MAE) dan terhadap performansi dengan menggunakan precision, recall dan accuracy. Berdasarkan hasil pengujian, didapatkan bahwa parameter n, γ, dan training set dengan test set mempengaruhi akurasi prediksi dan performansi dari recommender system menggunakan metode PEBE-CN maupun UBPS. Didapatkan pula bahwa akurasi prediksi dan performansi metode PEBE-CN terbukti lebih baik daripada metode UBPS.Kata Kunci : recommender system, collaborative filtering, UBPS, pearson, PEBE-CN, prediction errorABSTRACT: Recommender system with Collaborative Filtering methods somewhat inaccurate due to data sparsity. One of these methods for increase the precision accuracy of the calculation is using Prediction Error-Based Enhancement with Counting Number of Common Neighbors (PEBE-CN) Method. This method is the development of user-based collaborative filtering, User-Based Pearson Similarity (UBPS). This method predicted by looking prediction error of the items that have been rated by the active user, and weighted based on common neighbors of desired items. Several things are examined here which is the influence of the parameter n, γ, and comparison of the amount of training set and test set towards prediction accuracy PEBE-CN method by using Mean Absolute Error (MAE) and its perfomance with precision, recall and accuracy. Based on test results, it was found that the parameter n,γ, and comparison of the amount of training set and test set affects the accuracy of prediction and perfomance of recommender system with PEBE-CN and UBPS methods. Found also that the accuracy of prediction and it’s performance with PEBE-CN method proved to be better than UBPS method.Keyword: recommender system, collaborative filtering, UBPS, pearson, PEBE-CN, prediction error