Recommendations of Thesis Supervisor using the Cosine Similarity Method

Abstract

Selama proses penulisan skripsi, peran dosen pembimbing sangat dibutuhkan agar penyelesaiannya tepat waktu sehingga kualitas skripsi terjaga. Salah satu permasalahan dalam penentuan dosen pembimbing skripsi S1 ilmu komputer Universitas Bumigora adalah adanya unsur subyektifitas ketua prodi dalam penentuan dosen pembimbing skripsi. Tidak hanya itu, proses pemilihan dosen pembimbing dilakukan secara manual sehingga dapat memakan waktu lama dan dapat memperlambat pengerjaan skripsi mahasiswa. Pengerjaan skripsi mahasiswa menjadi terlambat dan tidak tepat waktu, jika topik skripsinya tidak sesuai dengan kompetensi dosen. Penelitian ini bertujuan implementasi metode cosine similarity untuk rekomendasi dosen pembimbing skripsi S1 ilmu komputer Universitas Bumigora. Tahapan penelitian ini terdiri dari pengumpulan Abstract During the thesis writing process, the role of the supervisor is needed so that the completion of the thesis is timely so that the quality of the thesis is maintained. One of the problems in determining the thesis supervisor for undergraduate computer science at Bumigora University is the subjectivity of the head of the study program in determining the thesis supervisor. Not only that, the selection process for supervisors is done manually so that it can take a long time and can slow down student thesis work. Students' thesis work will be late and not on time if the thesis topic is not in accordance with the competence of the lecturer. This study aims to apply the cosine similarity method to the recommendation of a thesis supervisor for undergraduate computer science at Bumigora University. The stages of this research consist of collecting thesis documents, pre-processing text (Case Folding, Tokenization, Filtering, Stemming), word weighting with TF-IDF, implementation of the cosine similarity method, and accuracy testing. The data used are 113 thesis documents which are divided as training data as many as 90 documents and testing data 23 documents. Based on the testing data on the test, the cosine similarity method can correctly recommend 21 of 23 thesis documents with an accuracy of 91.3%. Thus, the cosine similarity method can be applied to the case of selecting a thesis supervisor for undergraduate computer science at Bumigora University because it has very good accuracy.