Digitalisasi Pendidikan: Analisis Dan Prediksi Pola Pembelajaran Digital Pada LMS Edlink Menggunakan Algoritma Naive Bayes

Learning Management Systems (LMS) produce large volumes of learning activity data, yet at Universitas Lancang Kuning these data have so far been analysed only descriptively. This study builds a Naive Bayes model to classify and predict digital learning patterns on the Edlink LMS and designs an Early Warning System (EWS). Following the Knowledge Discovery in Databases framework, Edlink monitoring and evaluation reports for the odd and even semesters of 2025/2026 were extracted, cleaned, discretised, and labelled into three classes: Proactive, Standard, and Passive (At Risk). The dataset comprised 1,014 lecturer–study programme records (353 unique lecturers; 6,380 course classes). Categorical Naive Bayes with Laplace smoothing was evaluated using stratified 10-fold cross validation. Model 1 (within-semester classification) achieved a mean accuracy of 86.78%, with a recall of 95.11% and an F1-score of 95.47% for the Passive class. Model 2 (predicting the next semester, n = 333) achieved 73.27% accuracy and a Passive-class recall of 90.61%, comparable to a simple persistence rule (73.87%). Learning Assessment and Lesson Planning were the most influential variables. In the high-risk EWS group, 92.81% of lecturers were indeed Passive in the following semester. Edlink use increased significantly between semesters (Wilcoxon p < 0.001), although 63.98% of even-semester records remained Passive. Keyword: digitalisation of education; digital learning patterns; Naive Bayes; educational data mining; early warning system. Abstrak: Learning Management System (LMS) menghasilkan jejak data aktivitas pembelajaran dalam jumlah besar, tetapi di Universitas Lancang Kuning data tersebut selama ini hanya dianalisis secara deskriptif. Penelitian ini membangun model Naive Bayes untuk mengklasifikasikan dan memprediksi pola pembelajaran digital pada LMS Edlink serta merancang Early Warning System (EWS). Mengikuti kerangka Knowledge Discovery in Databases, laporan Monitoring dan Evaluasi Edlink Semester Ganjil dan Genap 2025/2026 diekstraksi, dibersihkan, didiskretisasi, dan dilabeli ke dalam tiga kelas: Proaktif, Standar, dan Pasif (Berisiko). Dataset terdiri atas 1.014 rekaman dosen–program studi (353 dosen unik; 6.380 kelas mata kuliah). Categorical Naive Bayes dengan Laplace smoothing dievaluasi menggunakan stratified 10-fold cross validation. Model 1 (klasifikasi dalam semester) mencapai akurasi rata-rata 86,78%, dengan recall 95,11% dan F1-score 95,47% pada kelas Pasif. Model 2 (prediksi semester berikutnya, n = 333) mencapai akurasi 73,27% dan recall kelas Pasif 90,61%, setara dengan aturan sederhana “kelas tetap” (73,87%). Learning Assessment dan Lesson Planning merupakan variabel paling berpengaruh. Pada kelompok risiko Tinggi EWS, 92,81% dosen benar-benar berada pada kelas Pasif di semester berikutnya. Penggunaan Edlink meningkat signifikan antarsemester (uji Wilcoxon p < 0,001), tetapi 63,98% rekaman Semester Genap masih berada pada kelas Pasif. Kata kunci: digitalisasi pendidikan; pola pembelajaran digital; Naive Bayes; educational data mining; early warning system.