Sistem kehadiran tradisional tidak mampu memprediksi risiko ketidakhadiran untuk mendukung intervensi dini sehingga sistem tersebut hanya bersifat reaktif. Penelitian ini bertujuan mengembangkan Sistem Peringatan Dini Kehadiran yang mengintegrasikan sistem kehadiran RFID dengan algoritma pembelajaran mesin untuk memprediksi kehadiran secara proaktif. Keunikan penelitian ini terletak pada penyusunan skema rekayasa fitur yang mengumpulkan 15 fitur prediktif dari data kehadiran serta penerapan model Random Forest untuk menghasilkan probabilitas risiko ketidakhadiran sebagai dasar peringatan dini. Penelitian ini menggunakan metodologi CRISP-DM (Proses Standar Antar industri untuk Penambangan Data) yang terdiri atas enam tahap, yaitu pemahaman bisnis, pemahaman data, persiapan data, pemodelan, evaluasi, dan implementasi. Model Random Forest yang diimplementasikan mencapai akurasi 87,3%, presisi 85,6%, recall 89,2%, dan skor F1 87,4% pada dataset yang terdiri dari 12.450 catatan. Fitur absence_last_7 days dan consecutive_absence menjadi prediktor paling berpengaruh dalam model. Berdasarkan probabilitas risiko yang dihasilkan model, sistem peringatan dini mampu mengidentifikasi 78% siswa berisiko tinggi satu minggu sebelum ketidakhadiran terjadi dan menurunkan tingkat ketidakhadiran sebesar 23% berdasarkan uji lapangan selama tiga bulan. Integrasi pembelajaran mesin meningkatkan kemampuan prediktif sistem kehadiran RFID secara signifikan sehingga mendukung pelaksanaan intervensi proaktif. Traditional attendance systems are unable to predict absenteeism risk to support early intervention; therefore, such systems remain reactive in nature. This study aims to develop an Early Warning Attendance System that integrates an RFID-based attendance system with machine learning algorithms to predict attendance proactively. The novelty of this study lies in the development of a feature engineering scheme that extracts 15 predictive features from attendance data, as well as in the implementation of a Random Forest model to generate absenteeism risk probabilities as the basis for early warning. This study employs the CRISP-DM (Cross-Industry Standard Process for Data Mining) methodology, which consists of six stages: business understanding, data understanding, data preparation, modeling, evaluation, and deployment. The implemented Random Forest model achieved an accuracy of 87.3%, precision of 85.6%, recall of 89.2%, and an F1-score of 87.4% on a dataset comprising 12,450 records. The features absence_last_7days and consecutive_absence were identified as the most influential predictors in the model. Based on the risk probabilities generated by the model, the early warning system was able to identify 78% of high-risk students one week before absenteeism occurred and reduce the absenteeism rate by 23% based on a three-month field trial. The integration of machine learning significantly improves the predictive capability of RFID-based attendance systems, thereby supporting the implementation of proactive interventions.
Affiliations come from OpenAlex; journal OAI-PMH metadata carries none. Authors without an OpenAlex ID are matched by name, which can merge or split people.