ABSTRAKSI: Preprocessing data adalah proses pembersihan data dari noise, outlier, missing value, irrelevant feature, redundant feature dll agar data yang digunakan bisa lebih efisien saat dilakukan proses data mining selanjutnya. Salah satu cara preprocessing adalah dengan mereduksi dimensi data. Metode reduksi data yang bisa dilakukan adalah dengan memilih feature-feature yang penting saja (feature selection). Pada tugas akhir ini diterapkan metode feature selection MMIMRSC (Maintaining mutual information and minimizing redundancy-synergy coefficient) yang bertujuan menghilangkan feature irelevan dan redundan dengan menghitung nilai informasi feature serta mengurangi redundansi feature dengan menghitung redundancy-synergy coefficientnya. Hasil performansi data hasil preprocessing, yaitu nilai precision dan recall nya mengalami peningkatan dan waktu pembentukan model klasifikasi menjadi lebih cepat.Kata Kunci : feature selection, mutual information, reduksi data, preprocessingABSTRACT: Preprocessing data is a process to cleaning data from noise, outlier, missing value, irrelevant feature, redundant feature etc, so that the data can be more efficient when the next mining process implemented. One of preprocessing techniques is by reduction the dimension of data. Data reduction method that can be implemented is feature selection (choosing only the important features). In this TA, feature selection method that being implemented is MMIMRSC (Maintaining mutual information and minimizing redundancy-synergy coefficient) that purpose to discard irrelevant and redundant feature by calculating the information value and also decreasing the redundant feature by calculating redundancy-synergy coefficient. The performance of reduces data after preprocessing shows that precision and recall value increase and time to build classification model become faster than unreduced data.Keyword: feature selection, mutual information, data reduction, preprocessing