SIMATUPANG, ANTONIO PARLINDUNGAN (2026) ANALISIS POLA KARIER ALUMNI MENGGUNAKAN ALGORITMA K-MEANS, HIERARCHICAL CLUSTERING, DAN DBSCAN BERDASARKAN DATA TRACER STUDY UNIVERSITAS MERCU BUANA. S1 thesis, Universitas Mercu Buana Jakarta.
|
Text (HAL COVER)
01 Cover.pdf Download (581kB) | Preview |
|
![[img]](https://repository.mercubuana.ac.id/style/images/fileicons/text.png) |
Text (BAB I)
02 Bab 1.pdf Restricted to Registered users only Download (167kB) |
|
|
Text (BAB II)
03 Bab 2.pdf Restricted to Registered users only Download (272kB) |
|
|
Text (BAB IIII)
04 Bab 3.pdf Restricted to Registered users only Download (422kB) |
|
|
Text (BAB IV)
05 Bab 4.pdf Restricted to Registered users only Download (1MB) |
|
|
Text (BAB V)
06 Bab 5.pdf Restricted to Registered users only Download (136kB) |
|
|
Text (DAFTAR PUSTAKA)
07 Daftar Pustaka.pdf Restricted to Registered users only Download (99kB) |
|
|
Text (LAMPIRAN)
08 Lampiran.pdf Restricted to Registered users only Download (665kB) |
Abstract
Graduate tracer studies play an important role in evaluating higher education outcomes by providing information on graduates' transition into the labor market. However, tracer study data are commonly analyzed using descriptive statistics, which limits the identification of underlying graduate career patterns. This study aims to identify graduate career patterns using tracer study data from Universitas Mercu Buana by applying K-Means, Hierarchical Clustering, and DBSCAN algorithms, comparing their clustering performance, and implementing the selected model in a web-based application. The dataset was obtained from the graduate tracer study conducted at Universitas Mercu Buana covering the period 2020–2024. The initial dataset consisted of 4,261 graduate records with 82 attributes. After data selection, data cleaning, feature engineering, ordinal encoding, logarithmic transformation of the income variable, and feature standardization using StandardScaler, a total of 2,191 records were retained for clustering analysis. Cluster quality was evaluated using the Silhouette Score and the Davies–Bouldin Index (DBI). The experimental results indicate that Hierarchical Clustering achieved slightly better internal evaluation scores than KMeans. Nevertheless, K-Means was selected as the final model because it produced more balanced clusters, offered better interpretability, and was more suitable for deployment in the proposed application. The selected model successfully identified four graduate career patterns, namely Adaptive Career, Aligned Career, Delayed Career, and Progressive Career, which were differentiated based on job waiting time, income level, horizontal job–education alignment, vertical education–job alignment, and company scale. The selected clustering model was subsequently implemented in a Streamlit-based Career Pattern Identification System (CPIS) featuring individual prediction, batch prediction, career pattern profiling, result visualization, and model information. The findings demonstrate that clustering techniques can effectively identify graduate career patterns and provide meaningful insights to support tracer study analysis, graduate outcome evaluation, and datadriven decision-making in higher education institutions. KAta kunci: Graduate Tracer Study, Career Pattern Analysis, Clustering, KMeans, Hierarchical Clustering, DBSCAN. Tracer study merupakan salah satu instrumen yang digunakan perguruan tinggi untuk mengevaluasi keberhasilan lulusan dalam memasuki dunia kerja. Namun, hasil tracer study umumnya masih dimanfaatkan dalam bentuk statistik deskriptif sehingga belum mampu mengidentifikasi pola karier alumni secara lebih mendalam. Penelitian ini bertujuan untuk mengidentifikasi pola karier alumni berdasarkan data tracer study Universitas Mercu Buana menggunakan algoritma KMeans, Hierarchical Clustering, dan DBSCAN, membandingkan performa ketiga algoritma tersebut, serta mengimplementasikan model terbaik ke dalam sebuah aplikasi berbasis web. Dataset yang digunakan berasal dari data tracer study alumni Universitas Mercu Buana periode 2020–2024. Dataset awal terdiri atas 4.261 data alumni dengan 82 atribut. Setelah melalui proses seleksi, pembersihan data, feature engineering, ordinal encoding, transformasi logaritmik pada variabel pendapatan, dan standardisasi menggunakan StandardScaler, diperoleh 2.191 data alumni yang siap digunakan pada proses clustering. Evaluasi kualitas cluster dilakukan menggunakan Silhouette Score dan Davies–Bouldin Index (DBI). Hasil penelitian menunjukkan bahwa algoritma Hierarchical Clustering memperoleh nilai evaluasi sedikit lebih baik dibandingkan K-Means. Namun, K-Means dipilih sebagai model akhir karena menghasilkan pembagian cluster yang lebih seimbang, lebih mudah diinterpretasikan, serta lebih sesuai untuk kebutuhan implementasi pada aplikasi. Model K-Means berhasil mengidentifikasi empat pola karier alumni, yaitu Karier Adaptif, Karier Selaras, Karier Tertunda, dan Karier Progresif, yang dibedakan berdasarkan karakteristik masa tunggu memperoleh pekerjaan, pendapatan, keselarasan pekerjaan dengan bidang studi, kesesuaian tingkat pendidikan terhadap pekerjaan, dan skala perusahaan tempat alumni bekerja. Model clustering yang dihasilkan kemudian diimplementasikan ke dalam aplikasi Career Pattern Identification System (CPIS) berbasis Streamlit yang menyediakan fitur prediksi individu, prediksi batch, visualisasi distribusi hasil, profil pola karier, dan informasi model. Hasil penelitian menunjukkan bahwa metode clustering mampu mengidentifikasi pola karier alumni secara efektif serta berpotensi mendukung pengelolaan data tracer study sebagai dasar evaluasi lulusan dan pengambilan keputusan di lingkungan perguruan tinggi. Kata kunci: Tracer Study, Clustering, K-Means, Hierarchical Clustering, DBSCAN, Pola Karier Alumni.
Actions (login required)
 |
View Item |
