Article Per Year (5 Year)
p-Index From 2021 - 2026
0.92
P-Index
This Author published in this journals
All Journal Jurnal Masyarakat Informatika ILKOM Jurnal Ilmiah Scientific Journal of Informatics Advance Sustainable Science, Engineering and Technology (ASSET)
Satriawan Rasyid Purnama
Universitas Diponegoro
Chemistry Computer Science & IT Control & Systems Engineering Decision Sciences, Operations Research & Management Electrical & Electronics Engineering Engineering Industrial & Manufacturing Engineering

Published : 5 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 5 Documents
Search

 1 
An Efficient Bidirectional Gated Recurrent Unit Approach for Student Study Duration Modeling and Timely Graduation Forecasting Satriawan Rasyid Purnama; Henri Tantyoko; Etna Vianita
Jurnal Masyarakat Informatika Vol 16, No 2 (2025): November 2025
Publisher : Department of Informatics, Universitas Diponegoro

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.14710/jmasif.16.2.73275

Abstract

Delays in student graduation remain a persistent challenge in higher education, with approximately 28% of students requiring more than four years to complete their studies, exceeding the standard duration. This study addresses the issue by proposing a predictive model to estimate students’ graduation year using a Bidirectional Gated Recurrent Unit (BiGRU) neural network. The model is trained on a combination of academic and financial indicators, including Grade Point (GP) scores from the first to the fifth semester, cumulative Grade Point Average (GPA), and the single tuition fee tier (UKT). The integration of these features allows the model to learn temporal patterns in students’ academic progression and financial capacity. Empirical analysis reveals that students in the UKT 8 group consistently demonstrate superior academic performance, as evidenced by their higher average GPA across semesters, compared to students in lower UKT groups. The BiGRU model achieves a Mean Absolute Percentage Error (MAPE) of 9.5%, indicating high predictive accuracy. These findings highlight the potential of deep learning models, particularly BiGRU, in forecasting academic outcomes. Furthermore, the insights generated from this model can serve as a valuable tool for universities in formulating targeted academic interventions and policies aimed at promoting timely graduation and reducing dropout rates.
Separable Convolutional Hierarchical Decomposition for Lightweight Residential Load Forecasting in Smart Grids Satriawan Rasyid Purnama; Henri Tantyoko; Adi Wibowo; Yesaya Rudolf Susanto Widyanto
Advance Sustainable Science Engineering and Technology Vol. 8 No. 3 (2026): May - July
Publisher : Science and Technology Research Centre Universitas PGRI Semarang

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.26877/asset.v8i3.2869

Abstract

Residential load forecasting is essential for maintaining grid stability and energy management in smart grids. However, achieving accurate real-time forecasting under resource constraints remains challenging because Transformer and LSTM models can be computationally demanding, while lightweight linear models such as DLinear have limited modeling flexibility. This study investigates whether a hierarchical separable convolutional framework can provide accurate and efficient residential load forecasting. To address this, SeparableCLF, a lightweight hierarchical decomposition model using depthwise separable convolution, is proposed and evaluated on hourly OpenEI residential load data from 20 U.S. states (2012) at forecast horizons of 6, 12, 24, 48, and 96 h. Relative to DLinear, SeparableCLF reduced MAPE by 0.93, 1.54, and 0.79 percentage points at 24, 48, and 96 h, respectively while requiring substantially fewer parameters than Transformer and LSTM models.SeparableCLF achieved the lowest MAPE at 12, 24, and 48 h. DLinear achieved the lowest errors at 6 h, whereas LSTM achieved the lowest MAPE at 96 h; at 96 h, SeparableCLF retained the lowest MAE, MSE, and RMSE among the compared models, indicating suitability for real-time forecasting on smart meters and edge-based smart grid devices.
A Comparative Evaluation of Model-Based and SHAP-Based Feature Analysis for Robust Health Data Classification Indra Waspada; Satriawan Rasyid Purnama; Alwey Hakim; Alfonso Clement Sutantio
Scientific Journal of Informatics Vol. 13 No. 2: May 2026
Publisher : Universitas Negeri Semarang

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.15294/sji.v13i2.42406

Abstract

Purpose: Feature selection is one critical element of healthcare data classification, which directly affects predictive performance, model robustness, and interpretability. Nevertheless, traditional model-based feature importance methods are unstable in robustness and provide random or misleading results on high-dimensional and heterogeneous healthcare data. The purpose of this paper was to assess model-based and SHapley Additive exPlanations (SHAP) - based feature analysis on multimodal healthcare data classification in a comparative manner. Methods: This study employed a quantitative comparative experimental design using real Electronic Medical Records (EMR) data captured from primary care clinics. The dataset comprises 2,158 patient records, with numerical and textual features in a multimodal feature space. The analytical pipeline included data acquisition, preprocessing, multimodal feature integration, and model development using six supervised learning algorithms from ensemble-based and margin-based categories. The model-based feature importance was used with the SHAP-based feature importance. We examined the robustness of this method across different scenarios through systematic feature ablation using baseline, strong, and weak features. Result: Experimental results demonstrate that model-based feature importance exhibits unpredictable behavior when features are removed and is sometimes counterintuitive. On the other hand, feature importance based on SHAP is consistent and linearly proportional across the models under consideration. The highest macro F1-score was achieved by Extra Trees with SHAP-based strong features (0.824), exceeding the baseline (0.811). In robustness testing, SHAP-based weak-feature removal reduced the LightGBM F1-score from 0.764 to 0.575, suggesting a clearer distinction between informative and non-informative features. Overall, SHAP-based feature selection provided a more reliable and interpretable framework for multimodal healthcare classification. Novelty: Our experiment results demonstrate empirically that the SHAP-based feature importance is more robust and reliable than conventional model-based approaches when it comes to extracting features from multimodal medical records. This work shows that SHAP is more than a post hoc explanation by presenting it as an interpretable feature selection criterion guiding feature relevance analysis in healthcare machine learning.
Multi-Horizon Short-Term Residential Load Forecasting Using Decomposition-Based Linear Neural Network Henri Tantyoko; Satriawan Rasyid Purnama; Etna Vianita
Advance Sustainable Science Engineering and Technology Vol. 7 No. 3 (2025): May - July
Publisher : Science and Technology Research Centre Universitas PGRI Semarang

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.26877/asset.v7i3.2033

Abstract

Short-Term Load Forecasting is crucial for grid stability and real-time energy management, particularly in residential settings where consumption is highly volatile and influenced by behavioral and external factors. Traditional models struggle to capture complex, non-linear patterns. This study proposes a forecasting framework based on the DLinear model, which decomposes time series data into trend and seasonal components using a simple linear neural network architecture. Designed for multi-horizon forecasting, the model predicts electricity demand across several future time points simultaneously. Experimental results show that DLinear performs best at a 24-hour prediction length, achieving the lowest MSE of 41.58 and MAE of 5.11, indicating improved accuracy with longer horizons. These results confirm DLinear’s robustness and efficiency in modeling dynamic residential electricity consumption patterns.
Feature Importance–Driven Multimodal Learning for Medical Diagnosis Classification Using Clinical and Symptom Text Data Indra Waspada; Satriawan Rasyid Purnama; Alfonso Clement Sutantio; Alwey Hakim
ILKOM Jurnal Ilmiah Vol 18, No 2 (2026)
Publisher : Prodi Teknik Informatika FIK Universitas Muslim Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33096/ilkom.v18i2.3289.255-267

Abstract

The integration of structured clinical measurements and unstructured textual symptom descriptions poses persistent challenges for automated medical diagnosis, particularly due to feature heterogeneity and class imbalance in real-world outpatient data. This study proposes a feature importance–driven multimodal machine learning framework for multi-class medical diagnosis classification that jointly models numerical clinical attributes and free-text symptom narratives within a unified pipeline. Beyond overall performance comparison, the proposed approach systematically examines the interaction between model architecture and feature importance through three controlled configurations: base, strong-feature, and weak-feature settings. Six supervised learning algorithms are evaluated using stratified five-fold cross-validation and imbalance-aware metrics. The results show that feature importance–driven modeling yields strongly model-dependent performance characteristics. Bagging-based tree ensembles benefit most from strong-feature selection, with the Extra Trees classifier achieving the best overall performance, reaching a macro-averaged F1-score of 0.821, compared to 0.811 in the base configuration and 0.642 when only weak features are retained. Conversely, margin-based classifiers rely on distributed feature representations. The kernel-based Support Vector Machine performs poorly under strong-feature selection (F1-score 0.450) but achieves a substantially higher F1-score of 0.795 and a macro-averaged recall of 0.808 under the weak-feature configuration. Linear SVM demonstrates stable behavior across configurations, maintaining a macro-averaged F1-score between 0.798 and 0.810, while attaining the highest overall recall of 0.819. These findings indicate that feature importance should be treated as a model-aware analytical tool for aligning feature selection strategies with the inductive bias of the learning algorithm, supporting robust and clinically meaningful diagnostic classification.
Co-Authors Adi Wibowo Alfonso Clement Sutantio Alfonso Clement Sutantio Alwey Hakim Alwey Hakim Etna Vianita Etna Vianita Henri Tantyoko Indra Waspada Indra Waspada Yesaya Rudolf Susanto Widyanto