What question did this study set out to answer?

This research aims to enhance the evaluation of virtual museums by using machine learning techniques to predict user satisfaction and engagement.

June 12, 2026International Journal of Data and Network Science1 citationsOpen Access

A machine learning-enhanced virtual museum for preserving Jordanian cultural heritage: Predicting user satisfaction and engagement profiling

ATArar Al Tawil SHSiti Hazyanti Mohd HashimUniversiti Sains Malaysia IKIsraa Wahbi Kamal

Key Points

This research aims to enhance the evaluation of virtual museums by using machine learning techniques to predict user satisfaction and engagement.
Implemented a machine learning framework to assess user experience in a virtual museum of Jordanian cultural heritage.
Evaluated 150 university students using Likert-scale ratings across four experience dimensions: engagement, cultural understanding, navigation intuitiveness, and visual realism.
Applied six machine learning classifiers, with Logistic Regression yielding the highest accuracy of 84.67%.
Logistic Regression classifier achieved an accuracy of 84.67% and an F1-score of 0.7884 for predicting user satisfaction.
Cultural understanding, visual realism, and navigation were identified as strong predictors of user satisfaction, with feature importances of 0.189, 0.173, and 0.168 respectively.
K-Means clustering revealed three user profiles, with 27.3% identified as Passive Observers and the rest as Selective Explorers.

Abstract

Virtual reality (VR) has emerged as a powerful tool for cultural heritage preservation; however, most existing VR museum studies rely on descriptive statistics for evaluation, lacking the predictive depth that data-driven approaches can offer. This paper addresses this gap by integrating machine learning (ML) techniques into the evaluation of a VR museum dedicated to traditional Jordanian women's clothing, focusing on Bedouin garments from Wadi Rum, Ma'an, and Petra. Seven historically accurate 3D costumes were reconstructed using Blender based on ethnographic references and expert consultations, integrated into a navigable museum environment in Unity, and deployed on the Meta Quest 2 headset. A total of 150 university students participated in the evaluation, providing demographic data and Likert-scale ratings across four experience dimensions: engagement, cultural understanding, navigation intuitiveness, and visual realism. Six ML classifiers, K-Nearest Neighbors, Support Vector Machine, Decision Tree, Random Forest, Naive Bayes, and Logistic Regression, were trained using 5-fold stratified cross-validation to predict user satisfaction. Logistic Regression achieved the best performance (accuracy: 84.67%, F1-score: 0.7884), while Decision Tree performed the lowest (accuracy: 71.33%, F1-score: 0.5947). Random Forest feature importance analysis revealed that cultural understanding (0.189), visual realism (0.173), and navigation (0.168) were the strongest predictors, whereas demographic factors such as gender (0.023) and prior VR experience (0.034) had minimal influence. K-Means clustering identified three distinct user profiles: Passive Observers (27.3%), and two Selective Explorer groups (39.3% and 33.3%) with divergent interaction patterns. These results demonstrate that ML can effectively predict and profile user satisfaction in VR heritage environments, offering data-driven insights for designing adaptive and personalized virtual museums.

问 AI

Bookmark

View Full Paper