ABSTRAKSI: Missing value membuat data menjadi kurang bermakna dan inkonsisten. Oleh Karena itu perlu tahap awal atau preprocessing. Terdapat beberapa metode imputasi untuk menangani missing value tersebut misal Shell neighbor. Proses awal metode tersebut, melakukan normalisasi dataset yaitu memetakan nilai kedalam range min and max value. Setelah normalisasi, dilakukan perhitungan jarak terhadap missing record. Jumlah missing record tergantung missing value rate yang digunakan. Berdasarkan hasil perhitungan jarak dipilih k tetangga terdekat dan left and right neighbor. Kemudian mencari bobot left and right neighbor. Hasil perkalian bobot terhadap nilai atribut kelas merupakan nilai prediksi missing value. Nilai prediksi tersebut akan dievaluasi terhadap nilai asli menggunakan parameter root mean square error. Berdasarkan hasil pengujian semaikin besar k nearest neighbor dengan missing value rate maka nilai RMSE semakin besar. Hal itu dikarenakan semakin besar nilai k nearest neighbor maka hasil prediksi semakin tidak tepat.Kata Kunci : MVR,left and right neighbor,RMSE,BobotABSTRACT: Missing values made data was not meaning and multi meaning. So that it’s needed early steps or preprocessing. There are many imputation methods for example shell-neighbor. Early process of shell neighbor methods is normalization into minmax value. After that,impute distance to missing record. Sum of missing record is depends missing value rate of used. According distances result then is choosen k nearest neighbor include left and right neighbor. Then search weight left and right nearest neighbor. The result of multiply weight to attribute values are prediction value. That values will be valuated into original values with rot mean square error parameters. According result of experiment,the highest k nearest neighbor with missing value rate then RMSE value highest. Because of higher k nearest neighbor then prediction result is incorrect.Keyword: MVR,left and right neighbor,RMSE,weight