Jurnal Sisfo Vol. 06 No. 01 (2016) 147–156 is.its.ac.id/pubs/oajis/ Penerapan Deep Sentiment Analysis pada Angket Penilaian Terbuka Menggunakan K-Nearest Neighbor Jane Riany * , Mohammad Fajar, Musfirah Putri Lukman Teknik Informatika, STMIK KHARISMA Makassar Abstract Manually assessment and analysis process for open questionnaire requires high cost. Therefore, the aims of this study are to apply the deep sentiment analysis of the course assessment questionnaire using K-Nearest Neighbor (KNN) and to measure the level of accuracy. Data collected through literature study on a number of related research, interviews about how to process the questionnaire, and test the proposed deep sentiment analysis scheme using Ms. Excel. The implementation process was done with categorizing test data into three (3) deep categories, i.e: how to teach, lecture time, and expectation. The three categories are then processed deeply by firstly doing preprocessing without stemming process, weighting the word using TF-IDF, and calculating the degree of similarity between the training data and test data, determining the value of the coefficient (k) and making the classification and determination of whether data is positive or negative sentiment. Evaluation results show that the implementation of Deep Sentiment analysis of the questionnaire can improve accuracy. The proposed system is able to classify the test data into three (3) categories deep sentiment analysis, with the average success rate of finding information system (recall) is 95,6%, the average level of accuracy (precision) is 59,4%, and the level of harmonization of the both parameters (average f measure) is 73,3%. Keywords: Deep Sentiment Analysis, Questionnaire, TF-IDF, K-Nearest Neighbor, KNN Abstrak Proses penilaian dan analisis angket terbuka secara manual membutuhkan biaya yang tinggi. Olehnya itu, penelitian ini bertujuan untuk menerapkan deep sentiment analysis pada angket penilaian terbuka menggunakan K Nearest Neighbor (KNN). Data angket yang digunakan yaitu angket penilaian perkuliahan terhadap dosen di STMIK KHARISMA Makassar yang diisi oleh setiap mahasiswa diakhir semester. Pengumpulan data dilakukan melalui wawancara dan uji coba deep sentiment analysis menggunakan Ms.Excel. Proses penerapan deep sentiment analysis dilakukan dengan mengkategorikan data uji kedalam tiga kategori yaitu: cara mengajar, waktu perkuliahan, dan harapan. Selanjutnya kategori tersebut diproses secara mendalam (deep) dengan melakukan preprocessing tanpa stemming, pembobotan kata menggunakan Term Frequence–Inverse Document Frequence, menghitung tingkat kemiripan antara data latih dan data uji, menentukan koefisien dan melakukan klasifikasi serta penetuan apakah data bermakna positif atau negatif. Evaluasi menunjukkan sistem mampu mengklasifikasikan data uji ke dalam tiga kategori deep sentiment analysis dengan hasil pengujian rata-rata tingkat keberhasilan sistem menemukan informasi sebesar 95,6%, rata-rata tingkat ketepatan sebesar 59,4%, dan tingkat harmonisasi keduanya sebesar 73,3%. Kata kunci: Deep Sentiment Analysis, Angket Penilaian, TF-IDF, K-Nearest Neighbor, KNN © 2016 Jurnal SISFO. Histori Artikel : Disubmit 25 Juli 2016; Diterima 23 September 2016; Tersedia online 25 September 2016 * Corresponding Author Email address: jane_13@kharisma.ac.id (Jane Riany)