Pengembangan Model Rekomendasi untuk Mendukung Keputusan Human Resource Berbasis Analisis Pola Kehadiran Pegawai dengan Pendekatan Data Analytics

Kottama, Bryan Agam (2026) Pengembangan Model Rekomendasi untuk Mendukung Keputusan Human Resource Berbasis Analisis Pola Kehadiran Pegawai dengan Pendekatan Data Analytics. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6032241150-Master_Thesis.pdf Restricted to Repository staff only Download (18MB) | Request a copy

Abstract

Kehadiran pegawai merupakan indikator penting kedisiplinan dan komitmen kerja pada institusi sektor publik. Di perguruan tinggi negeri, data presensi elektronik telah terkumpul dalam jumlah besar, namun pemanfaatannya masih dominan untuk keperluan administratif sehingga belum mendukung pengambilan keputusan sumber daya manusia (SDM) secara proaktif. Permasalahan penelitian adalah belum tersedianya model analitik terintegrasi yang mengolah data presensi menjadi informasi perilaku kehadiran yang dapat ditindaklanjuti unit SDM. Penelitian ini bertujuan mengembangkan model rekomendasi berbasis analisis pola kehadiran pegawai dengan pendekatan pembelajaran tanpa pengawasan (unsupervised learning) serta prototipe sistem pendukung keputusan (Decision Support System/DSS).
Penelitian dilakukan secara kuantitatif eksploratif pada sebuah perguruan tinggi negeri dengan memanfaatkan data presensi 2.151 pegawai aktif, meliputi dosen dan tenaga kependidikan (tendik), periode 2024-2025. Data diolah menjadi 54.543 observasi scorecard perilaku bulanan melalui pembersihan data, rekayasa fitur, analisis eksploratori data (Exploratory Data Analysis/EDA), clustering K-Means (K=4) dan DBSCAN pada populasi terpisah, serta evaluasi koefisien Silhouette dan Davies-Bouldin Index. Hasil segmentasi dipetakan ke label rekomendasi manajerial SDM dan divisualisasikan melalui dasbor DSS periodik.
Hasil penelitian menunjukkan empat segmen perilaku pada masing-masing populasi. Pada dosen, segmen Ketepatan Waktu Rendah mendominasi 72,89% observasi scorecard dengan median ketepatan waktu populasi 5,3%, yang mencerminkan masalah ketepatan waktu administratif bukan ketidakhadiran massal. Pada tendik, segmen Disiplin Tinggi mencakup 74,25% observasi disertai pemisahan tajam segmen risiko alpha. Algoritma DBSCAN melengkapi K-Means dengan mendeteksi deviasi episodik; 92,5% anomali dosen selaras aturan operasional SDM. Prototipe dasbor DSS mendukung audit skor, prioritas risiko, dan perbandingan antar unit sebelum keputusan disiplin. Integrasi scorecard perilaku, clustering, deteksi anomali, dan DSS terbukti memperkuat keputusan SDM berbasis bukti. Model yang dikembangkan merupakan alat analitik pendukung dan bukan pengganti proses hukum disiplin aparatur sipil negara.
=======================================================================================================================================
Employee attendance is an important indicator of workplace discipline and commitment in public institutions. At state universities, large volumes of electronic attendance data have been collected, yet their use remains largely administrative and does not yet support proactive human resource (HR) decision-making. The research problem is the absence of an integrated analytical model that transforms attendance data into actionable attendance-behavior information for HR units. This study aims to develop a recommendation model based on attendance pattern analysis using unsupervised learning and a Decision Support System (DSS) prototype. The research employed an exploratory quantitative approach at a state university using attendance records of 2,151 active employees, including lecturers and educational staff, from 2024 to 2025. The data were processed into 54,543 monthly behavioral scorecard observations through data cleaning, feature engineering, Exploratory Data Analysis (EDA), K-Means (K=4) and DBSCAN clustering on separate populations, and evaluation using Silhouette Coefficient and Davies-Bouldin Index. Segmentation outputs were mapped to managerial HR labels and visualized in a periodic DSS dasbor. The results show four behavioral segments in each population. Among lecturers, the combined Low Punctuality segments dominate 72.89% of scorecard observations, with a population median punctuality of 5.3%, reflecting an administrative punctuality issue rather than mass absenteeism. Among educational staff, the High Discipline segment covers 74.25% of observations with a distinct alpha-risk subgroup. DBSCAN complements K-Means by detecting episodic deviations; 92.5% of lecturer anomalies align with operational HR rules. The DSS prototype supports score auditing, risk prioritization, and cross-unit comparison before disciplinary decisions. The integration of behavioral scorecards, clustering, anomaly detection, and DSS strengthens evidence-based HR decisions. The developed model serves as a supportive analytical tool and does not replace formal civil servant disciplinary procedures.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Analitik SDM, Clustering, DBSCAN, Kehadiran pegawai, Sistem pendukung keputusan, Behavioral scorecard, Clustering, DBSCAN, Decision support system, Employee attendance
Subjects:	T Technology > T Technology (General) > T58.62 Decision support systems
Divisions:	Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT)
Depositing User:	Bryan Agam Kottama
Date Deposited:	29 Jul 2026 02:16
Last Modified:	29 Jul 2026 02:16
URI:	http://repository.its.ac.id/id/eprint/138716

Actions (login required)

View Item