Alokasi Dosen Pembimbing Pertama Tugas Akhir Mahasiswa dengan Pendekatan Cosine Similarity dan Genetic Algorithm

Abstract

Proses alokasi dosen pembimbing pertama tugas akhir mahasiswa di perguruan tinggi sering menghadapi kendala efisiensi karena perlu menyelaraskan kapasitas maksimum bimbingan dosen, dosen pilihan mahasiswa, dan kesesuaian topik penelitian berbasis teks terbuka. Penelitian ini bertujuan membangun mekanisme optimasi alokasi dosen pembimbing pertama menggunakan genetic algorithm (GA) yang dikombinasikan dengan cosine similarity untuk mengukur kemiripan topik. Permasalahan dimodelkan menggunakan metode penalti yang mengintegrasikan hard constraint (kapasitas maksimum bimbingan dosen) dan soft constraint (dosen pilihan mahasiswa dan topik penelitian). Data teks dibobotkan menggunakan term frequency-inverse document frequency (TF-IDF) dengan tiga skema n-gram (unigram, bigram, dan unibigram). GA kemudian dijalankan pada delapan skenario kombinasi jumlah generasi, mekanisme penalti kapasitas, dan skema pembobotan. Hasil penelitian menunjukkan bahwa hanya peningkatan jumlah generasi yang meningkatkan nilai fitness solusi dan waktu komputasi. S6- unibigram dipilih karena menghasilkan alokasi yang terbaik tanpa melanggar kapasitas maksimum bimbingan dosen, meminimumkan ketidaksesuaian dosen pilihan dan topik penelitian, serta efisien secara komputasi. Mekanisme terbaik ini kemudian diimplementasikan ke dalam dashboard interaktif berbasis R Shiny untuk mendukung pengambilan keputusan ke depannya.

The allocation process of the first supervisor for students' final assignments in higher education often faces efficiency challenges due to the need to align lecturer capacities, students' preferred lecturers, and research topic suitability based on open text. This study aims to build an optimization mechanism for allocating the first supervisor using a genetic algorithm (GA) combined with cosine similarity to measure topic similarity. The problem is modeled using a penalty method that integrates hard constraint (maximum advising capacity) and soft constraint (preferred lecturers and research topics). Text data are weighted using term frequency-inverse document frequency (TF-IDF) across three n-gram schemes (unigram, bigram, and unibigram). GA is then run across eight scenarios combining numbers of generations, capacity penalty mechanisms, and weighting scheme. The results show that only the increase in the number of generations improves the solution's fitness value and computation time. S6-unibigram was selected because it produced the best allocation without violating lecturers' maximum advising capacities, minimizing mismatches in preferred lecturers and research topics, and remaining computationally efficient. This best mechanism was then implemented into an interactive dashboard based on R Shiny to support decision-making in the future.