Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Triyanna Widiyaningtyas has a total of 164 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Meta-level based recommender system using knowledge graph-based neural collaborative filtering

    by Erfan Ainul Yakin, Triyanna Widiyaningtyas, Hary Suswanto, Didik Dwi Prasetya, Wahyu Caesarendra

  • Open Access

    Representation over Kernel: A Cross-Category Analysis of Semantic Review Embedding versus Rating-Overlap Collaborative Filtering for User Cold-Start

    by Moh Rahmat Irjii Matdoan, Triyanna Widiyaningtyas, Firmansyah Ibrahim, Rasna

    Purpose – The user cold-start problem cripples recommenders when a new user has almost no interaction history. This study asks two questions prior work conflates: does semantic review embedding similarity outperform conventional rating-overlap collaborative filtering (CF) in cold-start, and does a magnitude-sensitive kernel over the embedding add anything beyond ordinary cosine similarity? Design/methods/approach – Under one identical protocol on two Amazon review domains of differing sparsity, RBF-SBERT (an RBF kernel over SBERT embeddings) is compared with cosine, Pearson, an XGBoost baseline, rating-overlap CF, and two popularity baselines. Cold-start users (one to three interactions) use a per-user sampled protocol; significance uses Wilcoxon and Cohen's d. Findings – In the sparse domain the embedding beats even the strongest simple baseline, item-count popularity, on precision (P@5 0.567 vs 0.252; d=0.65) and ranking (AUC 0.869 vs 0.821); in the denser domain that edge vanishes as popularity matches AUC and exceeds precision, so the advantage is governed by sparsity. Rating-overlap CF is weakest, collapsing to its mean-rating fallback. The RBF kernel is inert (≈ cosine, d=0.007–0.107) because the pooled magnitude it uses is nearly constant (CV 0.07–0.09) and tracks review length, not preference. Research implications/limitations – Cold-start gains arise from the representation, not the kernel; methods must be benchmarked against popularity under sparsity-aware protocols. Evaluation covers two Amazon categories under one sampled protocol. Originality/value – The study separates representation from kernel and precision from ranking quality, giving a reproducible account of when review-based embedding helps cold-start and why.

  • Open Access

    FTFPOS-IDF: A Fuzzy Rule-Based Thematic Term Weighting Scheme for Bloom's Taxonomy Question Classification

    by Sucipto Sucipto, Didik Dwi Prasetya, Triyanna Widiyaningtyas

    The increasing adoption of Artificial Intelligence (AI) in education has created a growing demand for automated and reliable assessment systems. Existing Bloom’s Taxonomy (BT) question classification approaches commonly rely on TF-IDF-based weighting schemes, which assign static term weights and often fail to capture the varying thematic importance of terms across cognitive levels. To address this limitation, this study proposes a novel Fuzzy Thematic Feature and Part-of-Speech Inverse Document Frequency (FTFPOS-IDF) weighting scheme that integrates fuzzy rule-based reasoning with Natural Language Processing (NLP) to dynamically assign thematic weights according to Bloom’s Taxonomy relevance. The proposed framework combines Machine Learning (ML) and Deep Learning (DL) classifiers with Chi-Square feature selection to reduce irrelevant features and improve classification performance. Experimental results demonstrate that FTFPOS-IDF consistently outperforms conventional TF-IDF variants across multiple classification models. The highest performance was achieved by the Multilayer Perceptron (MLP) classifier with an accuracy of 86.7%. These findings indicate that fuzzy rule-based thematic weighting can effectively enhance Bloom’s Taxonomy question classification and support scalable, reliable, and sustainable digital assessment systems in educational environments.

  • Open Access

    SIEmo-LSTM: Multimodal Fusion of Text, Unicode Emoji, and ASCII Emoticons for Indonesian Sentiment Analysis

    by M. Noer Fadli Hidayat, Didik Dwi Prasetya, Triyanna Widiyaningtyas

    Sentiment analysis of Indonesian user reviews is challenged by informal language and frequent nonverbal cues such as Unicode emojis and ASCII emoticons. This study aimed to quantify the benefit of explicitly modeling ASCII emoticons together with Unicode emojis and text for three-class sentiment classification (negative/neutral/positive). SIEmo-LSTM is a tri-modal pipeline that (i) maps Unicode emojis using an emoji sentiment resource, (ii) detects, normalizes, and converts ASCII emoticons into descriptive tokens using Emot and LEED (93.3% successful conversion), and (iii) encodes the unified sequence using IndoBERT as a contextual feature extractor and refines it with a Bi-LSTM layer before multiclass prediction. Experiments used 304,570 Ruangguru app reviews (2016–2023), a tri-modal subset of 2,527 reviews, and a 70/20/10 train/validation/test split. Class imbalance was addressed using Random OverSampling (ROS). The full Text+SE+IE configuration with ROS achieved up to 0.9935 Accuracy and 0.9967 Macro-F1, outperforming text-only and text+Unicode baselines, while Random UnderSampling (RUS) consistently degraded performance. These findings imply that treating ASCII emoticons as a first-class affective modality—alongside Unicode emojis and text—improves robustness and class-balanced sentiment recognition for Indonesian user-generated reviews.


  • Open Access

    Fraud Identification in Online Financial Transactions Combining SMOTE and Ensemble Classifiers

    by Juvinal Ximenes Guterres, Triyanna Widiyaningtyas, Ilham Ari Elbaith Zaeni

    The increasing reliance on digital payment systems has intensified the exposure of financial institutions to credit card fraud, particularly under conditions of extreme class imbalance where fraudulent transactions account for less than 0.2% of total records. This study presents SMOVO, a Synthetic Minority Oversampling Technique (SMOTE)-based Voting Ensemble. It is a fraud detection framework designed to address this imbalance through the integration of three complementary components: Analysis of Variance (ANOVA) F-test-based feature selection, SMOTE-based data resampling, and a heterogeneous soft-voting ensemble comprising Random Forest (RF), XGBoost, LightGBM, and AdaBoost. Evaluated on the European Credit Card Fraud Dataset, SMOVO attained a recall of 0.962, F1-score of 0.864, Area Under the Curve (AUC) of 0.982, Matthews Correlation Coefficient (MCC) of 0.868, and Geometric mean (G-mean) of 0.980. In comparison with generative oversampling approaches such as ESMOTE-GAN, SMOVO exhibits more stable training behavior, reduced computational overhead, and improved model interpretability. These results indicate that SMOVO provides an effective and practically deployable solution for fraud detection in highly imbalanced financial transaction environments.


  • Open Access

    SELI: The Stacking-Blending Ensemble Learning and Interpretability Framework for Software Effort Estimation

    by Puguh Jayadi, Didik Dwi Prasetya, Triyanna Widiyaningtyas, Azlan Mohd Zain

    Software Effort Estimation (SEE) is an important element in software engineering that predicts the time, cost, and human resources needed for project development. Accurate estimates play a vital role in project planning and risk control, but many existing models still struggle to balance prediction accuracy and interpretability. Machine learning and deep learning–based approaches, such as SENSE, NIVIM, Mdb+MoWE, and GGSNN, can improve accuracy, but they are complex and challenging to explain. This study proposes the Stacking-Blending Ensemble Learning and Interpretability Framework (SELI), a lightweight, adaptive SEE framework integrated with Explainable Artificial Intelligence (XAI) to address these issues. SELI uses ridge-based meta-learning and non-negative blending to improve stability across datasets and integrates interpretability methods, such as Shapley Additive Explanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME), to provide insight into feature contributions. The results of experiments on six benchmark datasets show that SELI achieves Mean Absolute Error (MAE) values of 0.0042–0.0577 and coefficients of determination (R2) of up to 0.9939, outperforming the baseline model. XAI analysis reveals that SELI can provide consistent and transparent explanations of the factors influencing estimates. These findings indicate that SELI is an important step in developing an accurate SEE model and has great potential to support decision-making in software project management.


  • Open Access

    TOWARD A THEORY OF MULTIFORMAT SEMANTIC EMBEDDING FOR SMART CITIES: A SYSTEMATIC LITERATURE REVIEW ON JOINT TEXT AND INFOGRAPHIC REPRESENTATIONS IN BIG DATA GOVERNANCE

    by Endang Sri Sumanti, Didik Dwi Prasetya, Triyanna Widiyaningtyas

    Smart city ecosystems generate diverse data streams, including legal documents, textual citizen reports, sensor dashboards, geospatial maps, and infographics. Despite advances in natural language processing (NLP) embedding, vision-language models, and graph-based urban analytics, a coherent theoretical framework explaining the integration of semantic representations across multiple modalities into a single embedding space for decision- making is lacking. This paper describes a systematic literature review (SLR) pertaining to multi-format semantic embedding in smart cities, emphasizing the combined representation of textual documents and visual artifacts such as infographics (maps, dashboards, plots, and schematic diagrams). Following PRISMA 2020 guidelines, we conducted searches in SpringerLink, Taylor & Francis, and SCOPUS between 2021 and 2025, resulting in 2,662 records and the inclusion of 110 primary studies through a screening process. This state of the art research is based on eight questions addressing data modalities, urban application domains, embedding models, integration mechanisms, evaluation metrics, and governance considerations. The review explains that most studies continue to model each modality separately, cross-modal integration is largely characterized by fusion-based methodologies and knowledge graphs, and evaluations emphasize retrieval and prediction accuracy, with insufficient focus on semantic coherence and interpretability at the policy level. Given these identified gaps, we propose the Theory of Multiformat Semantic Integration (TISM) and the Urban Multiformat Embedding Framework (UMEF), alongside a conceptual metric called Urban Semantic Coherence (USC) to evaluate cross-modal alignment. The review concludes with a proposed research agenda aimed at fostering robust, explainable, and governable multiformat embedding in the context of smart city big data environments.

  • Open Access

    HCF-MFGB: Hybrid Collaborative Filtering Based on Matrix Factorization and Gradient Boosting

    by Salahudin Robo, Triyanna Widiyaningtyas, Wahyu Sakti Gunawan Irianto

    Recommendation systems are an integral and indispensable part of every digital platform, as they can suggest content or items to users based on their respective needs. Collaborative filtering is a technique often used in various studies, which produces recommendations by analyzing similarities between users and items based on their behavior. Although often used, traditional collaborative filtering techniques still face the main challenge of sparsity. Sparsity problems occur when the data in the system is sparse, meaning that only a portion of users provide feedback on some items, resulting in inaccurate recommendations generated by the system. To overcome this problem, we developed a Hybrid Collaborative Filtering model based on Matrix Factorization and Gradient Boosting (HCF-MFGB), a new hybrid approach. Our proposed model integrates SVD++, the XGBoost ensemble learning algorithm, and utilizes user demographic data and meta items. We utilize information, both explicitly and implicitly, to learn user preference patterns using SVD++. The XGBoost algorithm is used to create hundreds of decision trees incrementally, thereby improving model accuracy. Meanwhile, user demographic and meta-item data are clustered using the K-Means Clustering algorithm to capture similarities in user and item characteristics. This combination is designed to improve rating prediction accuracy by reducing reliance on minimal explicit rating data, while addressing sparsity issues in movie recommendation systems. The results of experiments on the MovieLens 100K, MovieLens 1M, and CiaoDVD datasets show significant improvements, outperforming various other baseline models in terms of RMSE and MAE. On the MovieLens 100K dataset, the HCF-MFGB model obtained an RMSE value of 0.853 and an MAE value of 0.674. On the MovieLens 1M dataset, the HCF-MFGB model obtained an RMSE value of 0.763 and an MAE value of 0.61. On the CiaoDCD dataset, the HCF-MFGB model achieved an RMSE value of 0.718 and an MAE value of 0.495. These results confirm a significant improvement in movie recommendation accuracy with the proposed approach.


  • Open Access

    Model Graf Spasial Adaptif untuk Optimasi Jalur UAV pada Sistem Mitigasi Kebakaran Hutan

    by Adinda Salsa Leviona, Triyanna Widiyaningtyas

    Dynamic wildfires require adaptive mitigation systems capable of responding in real time. Optimizing Unmanned Aerial Vehicle (UAV) routes is critical for effective fire mitigation. This study compares the efficiency of the Dijkstra, A*(star), and Bidirectional Dijkstra algorithms on spatial hotspots graph. The quantitative experimental design used data-driven simulation. Data were obtained from the Forest Fire Area dataset (Kaggle), containing 517 wildfires with (X, Y) coordinates in the Montesinho Natural Park, Portugal. Preprocessing involved analyzing spatial patterns using scatter plots and de-duplication of coordinates. These coordinates were then utilized as nodes to construct a graph using Triangulasi Delaunay. The algorithms were implemented using custom-coded. Performance was evaluated through repeated tests based on the metrics of execution time, memory usage, nodes explored, and path length. The results show that all three algorithms achieved identical optimal paths, however A* achieved the highest computational efficiency by balancing time and memory usage. The combination of the Delaunay spatial graph and the A* algorithm effectively reduces search complexity, it suitable for adaptive UAV navigation. This study provides an empirical comparison of algorithms on a spatial hot-spots graph using Triangulation Delaunay, a method rarely used to evaluate the efficiency of UAV path tracing in real spatial conditions.

  • Open Access

    A Novel LBFGS Optimization Algorithm for RIS-Assisted MIMO Systems with Hardware Impairments

    by Soraya Norma Mustika, Muladi Muladi, Triyanna Widiyaningtyas

    Reconfigurable Intelligent Surfaces (RIS) have emerged as a promising technology to overcome the spectral efficiency limitations of conventional MIMO systems, particularly under real-world conditions characterized by hardware imperfections and non-ideal reflection properties. This paper proposes an innovative Phase and Reflection Optimization Algorithm (PROA) that simultaneously optimizes both the base station beamforming and RIS phase configurations through a computationally efficient dual-stage iterative approach. By combining Singular Value Decomposition (SVD) for precoding with an Limited-memory Broyden-Fletcher-Goldfarb-Shanno (L-BFGS) based phase optimization procedure that accounts for practical constraints such as amplitude-phase coupling, impedance variations, and imperfect channel state information (CSI), the proposed algorithm addresses the complex non-convex optimization problem inherent in RIS-assisted communications. The simulation results show that PROA achieves a spectral efficiency exceeding 80 bps/Hz while reducing the memory requirement by 80% and achieving speed improvements of up to three orders of magnitude compared with existing techniques. These results position PROA as a practical and high-performance solution for RIS deployment in 5G-Advanced and 6G networks.