Evaluasi komparatif model machine learning pada weak labeled behaviometrics dataset sistem pelatihan vr hospitality untuk anak down syndrome

DETAIL KOLEKSI

Evaluasi komparatif model machine learning pada weak labeled behaviometrics dataset sistem pelatihan vr hospitality untuk anak down syndrome

Oleh : Tarum Widyasti Pertiwi

Info Katalog

Penerbit : FTI - Usakti

Kota Terbit : Jakarta

Tahun Terbit : 2026

Pembimbing 1 : Dian Pratiwi

Pembimbing 2 : Anung Barlianto Ariwibowo

Kata Kunci : Machine Learning, Weak-Labeled Dataset, Behaviometrics, Virtual Reality, Random Forest, Down Syndrom

Status Posting : Published

Status : Lengkap

File Repositori

No.	Nama File	Hal.
1.	2026_SK_STF_064002200027_Halaman-Judul.pdf
2.	2026_SK_STF_064002200027_Surat-Pernyataan-Revisi-Terakhir.pdf	1
3.	2026_SK_STF_064002200027_Surat-Hasil-Similaritas.pdf	1
4.	2026_SK_STF_064002200027_Halaman-Pernyataan-Persetujuan-Publikasi-Tugas-Akhir-untuk-Kepentingan-Akademis.pdf	1
5.	2026_SK_STF_064002200027_Lembar-Pengesahan.pdf	1
6.	2026_SK_STF_064002200027_Pernyataan-Orisinalitas.pdf	1
7.	2026_SK_STF_064002200027_Formulir-Persetujuan-Publikasi-Karya-Ilmiah.pdf	1
8.	2026_SK_STF_064002200027_Bab-1-Pendahuluan.pdf
9.	2026_SK_STF_064002200027_Bab-2-Landasan-Teori.pdf
10.	2026_SK_STF_064002200027_Bab-3-Metodologi-Penelitian.pdf
11.	2026_SK_STF_064002200027_Bab-4-Analisis-dan-Pembahasan.pdf
12.	2026_SK_STF_064002200027_Bab-5-Kesimpulan-dan-Saran.pdf
13.	2026_SK_STF_064002200027_Daftar-Pustaka.pdf
14.	2026_SK_STF_064002200027_Lampiran.pdf

Abstrak
Abstract

P Pelatihan hospitality berbasis virtual reality (vr) menghasilkan behaviometrics dataset yang merekam perilaku pengguna selama mengikuti pelatihan. dataset yang digunakan dalam penelitian ini termasuk weak-labeled dataset karena proses pelabelannya menggunakan pendekatan rule-based labeling, dengan aturan pelabelan dan nilai threshold yang ditentukan berdasarkan analisis terhadap data hasil observasi dan penilaian pendamping anak down syndrome. penelitian ini bertujuan untuk membandingkan performa algoritma logistic regression, random forest, dan support vector machine (svm) dalam memprediksi status kelulusan peserta pelatihan berdasarkan weak-labeled behaviometrics dataset, serta mengevaluasi kemampuan generalisasi masing-masing model terhadap data baru. dataset pembangunan model terdiri atas 406 data sesi pelatihan yang diperoleh dari 20 mahasiswa dan 10 anak disabilitas, sedangkan pengujian kemampuan generalisasi dilakukan menggunakan 123 data sesi permainan yang berasal dari 27 anak down syndrome. evaluasi model dilakukan menggunakan metode stratified 5-fold cross validation dengan metrik accuracy, precision, recall, f1-score, balanced accuracy, roc-auc, dan confusion matrix, sedangkan kemampuan generalisasi dievaluasi berdasarkan tingkat kesesuaian hasil prediksi model terhadap penilaian pendamping sebagai ground truth. hasil penelitian menunjukkan bahwa random forest memberikan performa terbaik dengan f1-score sebesar 0,9797, balanced accuracy sebesar 0,9215, dan roc-auc sebesar 0,9926, serta memperoleh tingkat kesesuaian prediksi sebesar 88,62% pada pengujian menggunakan data baru. hasil tersebut menunjukkan bahwa random forest memiliki kemampuan klasifikasi dan generalisasi yang lebih baik dibandingkan logistic regression dan support vector machine, sehingga berpotensi mendukung proses penilaian kelulusan peserta pada sistem pelatihan hospitality berbasis virtual reality.

V Virtual reality (vr)-based hospitality training generates a behaviometrics dataset that records participants\\\' behavioral characteristics throughout the training sessions. the dataset used in this study is categorized as a weak-labeled dataset because the class labels were generated using a rule-based labeling approach, where the labeling rules and threshold values were determined based on the analysis of observational data assessed by companions of children with down syndrome. this study aims to compare the performance of logistic regression, random forest, and support vector machine (svm) in predicting participants\\\' training outcomes using a weak-labeled behaviometrics dataset and to evaluate the generalization capability of each model on unseen data. the model development dataset consisted of 406 training sessions collected from 20 university students and 10 children with disabilities, while the generalization test was conducted using 123 gameplay sessions from 27 children with down syndrome. model performance was evaluated using stratified 5-fold cross validation with accuracy, precision, recall, f1-score, balanced accuracy, roc-auc, and confusion matrix. the generalization capability was further assessed by comparing the model predictions with the companions\\\' evaluations, which served as the ground truth. the results indicate that random forest achieved the best performance, with an f1-score of 0.9797, balanced accuracy of 0.9215, and roc-auc of 0.9926. furthermore, random forest achieved the highest prediction agreement of 88.62% on the unseen gameplay data. these findings demonstrate that random forest outperformed logistic regression and support vector machine in both classification performance and generalization capability, making it the most reliable model for supporting participant assessment in virtual reality-based hospitality training.