Jurnal PTIPD: Penelitian Terapan Informatika dan Pengembangan Data

Implementasi Markerless Augmented Reality Berbasis Web untuk Simulasi Penataan Ruang Interior Menggunakan 8th Wall dan Three.js

2026-10-05T02:42:13+00:00

The interior design process often faces challenges when consumers struggle to visualize how furniture will look in their physical space before purchasing. Traditional Augmented Reality (AR) applications typically require native app installations, which creates accessibility barriers. This study implements a Web-based Augmented Reality (WebAR) application for interior design simulation using the 8th Wall engine and Three.js framework. Utilizing Simultaneous Localization and Mapping (SLAM) technology, the application enables markerless tracking, allowing users to project 3D furniture models directly onto their physical floors via a standard mobile web browser. The system features an integrated 3D model catalog and multi-gesture interactions, including single-finger drag for translation, two-finger pinch for scaling and rotation, and a raycasting-based tap mechanism for object selection and deletion. Black-box testing on various mobile browsers confirmed that the system successfully fulfilled 9 out of 10 designed functional test scenarios. While the SLAM tracking and gesture controls function accurately, the evaluation identified a technical limitation in the web-based wall collision detection. The findings demonstrate that WebAR provides a practical, installation-free alternative for immersive interior space planning.

Generative AI and Islamic Family Law: Normative Framing of Marital Relations

2026-09-11T10:42:28+00:00

This study examines how GPT frames authority, responsibility, reciprocity, and shared agency in controversial Islamic family law issues. The study used qualitative thematic analysis, descriptive quantitative mapping, and cross-question consistency analysis of 32 GPT responses obtained through independent sessions with standardized prompt protocols. The results showed that relational framing was the dominant orientation (20/32; 62.5%), followed by contextual (25%) and mixed (12.5%) framing, with no predominantly hierarchical responses. Qiwāmah tended to be framed as functional responsibilities that resulted in limited authority, rather than unlimited unilateral authority. Fixed alimony was primarily positioned as the husband's responsibility, but the wife's economic contribution was recognized in the practical family setting. Decision-making was increasingly geared towards consultation, participation, and negotiation. The findings resulted in the Relational Normative Reconciliation Model (RNRM), which links differentiated roles, functional responsibility, limited authority, reciprocal accountability, shared agency, and relational decisions. The research confirms that the normative coherence of AI outputs is not the same as legal validity, so human verification of Islamic sources and positive laws remains necessary.

Typological Analysis of Sultanate Tombs in Sumatra Using K-Means and Random Forest Clustering Methods

2026-10-05T02:42:27+00:00

The diversity of the morphological characteristics of Islamic tombs from the Sultanates in Sumatra reflects variations in cultural traditions, construction techniques, and architectural forms. However, the identification of tomb typologies is still dominated by manual observation; therefore, the classification process is relatively subjective. This study aimed to identify the typology of tombs from the Sultanates in Sumatra using the K-Means Clustering algorithm and to determine the most influential classification variables using the Random Forest algorithm. This study analyzed 85 tomb records consisting of four morphometric variables (tombstone height, length, width, and height) and several archaeological attributes. The Elbow method was used to determine the optimal number of clusters, resulting in three clusters with distinct morphometric characteristics. The evaluation of the clustering quality resulted in an average silhouette coefficient of 0.267, indicating a fairly good clustering structure. Furthermore, the Random Forest model achieved an accuracy of 68.75% with a kappa value of 0.521, indicating moderate classification performance. Variable importance analysis revealed that grave material, footstone shape, and tombstone shape were the most influential variables for distinguishing tomb typologies. These findings indicate that the combination of K-Means Clustering and Random Forest provides an effective and objective approach to identifying archaeological tomb typologies based on morphometric and physical attributes.

Implementation of a Customer Churn Model Using XGBoost and SHAP for Telecommunication Customers

2026-09-21T15:13:09+00:00

Customer churn is a condition in which customers stop using telecommunications services, potentially reducing revenue and customer loyalty while increasing customer acquisition costs. This study implements XGBoost with Synthetic Minority Over-sampling Technique (SMOTE) to address class imbalance and Shapley Additive exPlanations (SHAP) to interpret the factors contributing to churn predictions. The dataset used is the Telecom Customer Churn dataset from Maven Analytics, consisting of 7,043 customer records, with 6,589 records used after excluding customers with Joined status. The optimized XGBoost model achieved a Recall of 77.72%, an F1-score of 77.93%, and a ROC-AUC of 93.73%. SHAP analysis identified the factors contributing to customer churn predictions, while validation using Churn Category and Churn Reason showed that Competitor, Price, and Dissatisfaction accounted for 73.46% of churned customers. The most frequent churn reasons were Competitor had better devices (313 customers), Competitor made better offer (311 customers), and Attitude of support person (220 customers). The results provide analytical support for identifying customers likely to churn and understanding factors associated with churn predictions.

Design and Implementation of Marker-Based Augmented Reality Application for Waste Sorting Education

2026-10-05T02:54:28+00:00

Proper waste management starting from the household level is a critical step in pre-serving environmental ecosystems. However, community awareness and knowledge regarding waste sorting (Organic, Inorganic, and Hazardous/B3) remain low. Conven-tional educational tools such as brochures and static posters are often ineffective in en-gaging younger generations. This study presents EcoWaste-AR, a mobile Augmented Reality (AR) application developed using Unity 6 (6000.3.16f1) and Vuforia Engine SDK (v11.4.4) to provide an interactive and immersive waste sorting learning media. Following the Waterfall Software Development Life Cycle (SDLC) model, the system maps physical waste category cards (image markers) to interactive 3D assets: an ap-ple core for Organic waste, a plastic bottle for Inorganic waste, and a battery for B3 waste. Each model is optimized as low-poly meshes to ensure high frame rates (> 60 FPS) on standard Android devices. Interaction is enhanced by billboarding text labels for camera-facing legibility and smooth UI animations. Evaluation through black-box testing demonstrates successful marker tracking across varying viewing angles (30 to 90 degrees), distances (10 to 60 cm), and lighting conditions (> 10 lux). The resulting application serves as an engaging and technically robust tool to raise ecological literacy and foster waste sorting habits.