Claim Missing Document
Check
Articles

Found 6 Documents
Search
Journal : Kesatria : Jurnal Penerapan Sistem Informasi (Komputer dan Manajemen)

Synergistic Machine Learning: Enhancing Diabetes Prediction with Hybrid Deep Learning and Ensemble Models Airlangga, Gregorius
Kesatria : Jurnal Penerapan Sistem Informasi (Komputer dan Manajemen) Vol 5, No 3 (2024): Edisi Juli
Publisher : LPPM STIKOM Tunas Bangsa

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30645/kesatria.v5i3.457

Abstract

Diabetes, a growing global health concern, necessitates improved predictive strategies for early and accurate detection. This study evaluates the efficacy of various machine learning and deep learning models in predicting the onset of diabetes, employing a comprehensive dataset that includes clinical and demographic variables. Traditional machine learning models such as Decision Trees, Random Forest, KNN, and XGBoost provided foundational insights, with ensemble methods showing superior performance. Furthermore, we explored the potential of deep learning by analyzing a Simple Dense Neural Network (DNN), Convolutional Neural Network (CNN), and Recurrent Neural Network (RNN). While these individual models yielded valuable findings, particularly in identifying true positive cases, they did not surpass the ensemble techniques in overall accuracy. The pinnacle of our research was the development of a Deep Learning Meta Learner that combined Random Forest and Gradient Boosting predictions, achieving near-perfect classification metrics, and underscoring the strength of model integration. Our findings advocate for a hybrid predictive approach that merges the nuanced feature detection of deep learning with the robust pattern recognition of ensemble models, providing an impactful direction for future diabetes prediction research. This study contributes to the advancement of medical informatics and aims to support healthcare professionals in delivering proactive and personalized patient care.
Enhancing Concrete Compressive Strength Prediction with Deep Learning: A Comparative Analysis of Model Architectures Airlangga, Gregorius
Kesatria : Jurnal Penerapan Sistem Informasi (Komputer dan Manajemen) Vol 5, No 3 (2024): Edisi Juli
Publisher : LPPM STIKOM Tunas Bangsa

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30645/kesatria.v5i3.459

Abstract

The imperative to predict concrete compressive strength accurately is a crucial aspect of modern civil engineering, with significant implications for the safety and cost-effectiveness of construction projects. This research explores the application of deep learning techniques to enhance predictive accuracy in this domain. We conducted a comprehensive comparative analysis of five machine learning models: a Basic neural network model, a Dropout model, a Batch Normalization model, a Deep Dense Neural Network (Deep DNN), and a Convolutional Neural Network (CNN). Utilizing a dataset reflective of various concrete mixtures and their corresponding compressive strengths, each model underwent rigorous evaluation through a five-fold cross-validation scheme. Performance metrics, including Mean Squared Error (MSE) and R-Squared (R²), were computed to assess each model's predictive capabilities. The results indicated that models employing batch normalization and deeper architectures provided superior predictive performance, suggesting that these features are instrumental in understanding the complex relationships between the components of concrete mixtures. The Batch Normalization and Deep DNN models demonstrated remarkable accuracy and consistency, surpassing traditional and CNN models. This study not only enhances the current understanding of material property prediction through machine learning but also paves the way for the development of more efficient and robust predictive tools in civil engineering. The findings underscore the transformative potential of deep learning in material science, emphasizing its ability to deliver nuanced and precise predictions for critical engineering properties.
Hybrid Ensemble Model for Real-Time Intrusion Detection in IoT Networks Using Machine Learning and Deep Learning Techniques Airlangga, Gregorius
Kesatria : Jurnal Penerapan Sistem Informasi (Komputer dan Manajemen) Vol 5, No 4 (2024): Edisi Oktober
Publisher : LPPM STIKOM Tunas Bangsa

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30645/kesatria.v5i4.523

Abstract

The rapid growth of the Internet of Things (IoT) has introduced new security challenges, as IoT devices are increasingly vulnerable to sophisticated cyberattacks. This study proposes a hybrid ensemble model combining classical machine learning algorithms (Random Forest, Gradient Boosting) with deep learning (Multi-Layer Perceptron) to improve the detection of malicious activities in IoT networks. The model leverages the RT-IoT2022 dataset, which includes diverse attack patterns such as DDoS, Brute-Force SSH, and Nmap scanning. The integration of these models using a Voting Classifier achieves superior performance by exploiting the strengths of each individual model. Evaluation results demonstrate that the hybrid model outperforms its individual components, achieving an accuracy of 99.80%, precision of 99.80%, recall of 99.80%, and F1-score of 99.80%. The proposed system demonstrates strong generalization across both frequent and rare attack types, making it well-suited for real-world IoT environments where high accuracy and low false-positive rates are critical. This study contributes to the development of robust and scalable intrusion detection systems that can adapt to evolving threats in real-time
Comparative Analysis of Deep Learning Models for Predicting Fan Actuator Status in IoT-Enabled Smart Greenhouses Airlangga, Gregorius
Kesatria : Jurnal Penerapan Sistem Informasi (Komputer dan Manajemen) Vol 5, No 4 (2024): Edisi Oktober
Publisher : LPPM STIKOM Tunas Bangsa

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30645/kesatria.v5i4.524

Abstract

In this study, we propose a comprehensive comparison of deep learning models for predicting the status of fan actuators in an IoT-enabled smart greenhouse environment. The dataset, consisting of 37,923 observations, captures environmental variables such as temperature, humidity, and soil nutrient levels, alongside actuator statuses. The aim is to accurately predict the binary status of the fan actuator (on or off) based on these environmental conditions. To address the challenge of class imbalance in the dataset, we apply the Synthetic Minority Oversampling Technique (SMOTE) to generate synthetic samples of the minority class, ensuring a balanced distribution for training. Three deep learning architectures Multi-Layer Perceptron (MLP), Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM) are implemented and evaluated using 10-fold cross-validation. The performance of each model is assessed using accuracy, precision, recall, and F1 score metrics. Results indicate that all models demonstrate strong predictive capabilities, with the LSTM excelling in capturing temporal dependencies, the CNN effectively extracting spatial patterns, and the MLP achieving overall high accuracy in structured data. The findings of this study provide valuable insights into the strengths and weaknesses of these models for actuator status prediction, which can guide future developments in smart greenhouse automation systems
Machine Learning for Tsunami Prediction: A Comparative Analysis of Ensemble and Deep Learning Models Airlangga, Gregorius
Kesatria : Jurnal Penerapan Sistem Informasi (Komputer dan Manajemen) Vol 6, No 1 (2025): Edisi Januari
Publisher : LPPM STIKOM Tunas Bangsa

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30645/kesatria.v6i1.572

Abstract

Tsunamis, triggered by seismic activities, pose significant threats to coastal regions, necessitating accurate prediction models to mitigate their impact. This study explores the application of machine learning models, including ensemble methods (Random Forest, Gradient Boosting, XGBoost, LightGBM, and CatBoost) and deep learning (Neural Networks), for tsunami prediction based on seismic data. The dataset spans seismic events from 1995 to 2023, characterized by features such as magnitude, depth, and geographic location. A 10-fold cross-validation approach was employed to evaluate model performance using precision, recall, F1-score, accuracy, and ROC-AUC metrics. The results highlight that Gradient Boosting achieved the best balance between precision and recall, with an F1-score of 0.6544 and the highest ROC-AUC of 0.8606, demonstrating its strong discriminatory power. Random Forest excelled in precision (0.6920) and F1-score (0.6287), making it suitable for reducing false positives. Ensemble boosting models, such as CatBoost and LightGBM, offered consistent performance with low variability across folds. In contrast, Neural Networks underperformed, achieving an F1-score of 0.5497 and an ROC-AUC of 0.7936, indicating the need for further optimization. Despite promising results, challenges in recall scores underscore the need for enhanced detection of tsunami-triggering events. The findings establish ensemble methods, particularly Gradient Boosting and Random Forest, as robust tools for tsunami prediction, providing a foundation for early warning systems. Future work will focus on improving recall and exploring hybrid modeling techniques to optimize predictive accuracy and reliability.
Comparative Evaluation of CNN, LSTM, and GRU Architectures for Tsunami Prediction Using Seismic Data Airlangga, Gregorius
Kesatria : Jurnal Penerapan Sistem Informasi (Komputer dan Manajemen) Vol 6, No 1 (2025): Edisi Januari
Publisher : LPPM STIKOM Tunas Bangsa

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30645/kesatria.v6i1.573

Abstract

Tsunamis are among the most catastrophic natural disasters, often triggered by seismic events such as earthquakes. Accurately predicting tsunami occurrences based on seismic parameters is critical for mitigating their devastating impacts. This study investigates the application of three advanced deep learning architectures such as Convolutional Neural Networks (CNNs), Long Short-Term Memory Networks (LSTMs), and Gated Recurrent Units (GRUs) for binary classification of tsunami events using seismic data. The dataset comprises earthquake records from 1995 to 2023, including features such as magnitude, depth, latitude, longitude, Modified Mercalli Intensity (MMI), and Community Internet Intensity (CDI). The models were evaluated using stratified 10-fold cross-validation and assessed across precision, recall, F1-score, accuracy, and ROC-AUC metrics. Results indicate that CNN outperformed the other architectures, achieving the highest accuracy (72.5%), precision (0.5987), and ROC-AUC (0.7838). GRU demonstrated moderate performance, balancing computational efficiency and predictive accuracy with an accuracy of 71.7% and ROC-AUC of 0.7709. LSTM, while theoretically adept at modeling temporal dependencies, showed the lowest performance due to challenges in capturing the dataset’s characteristics. The findings emphasize the importance of selecting architecture suited to the dataset’s features and task requirements. CNN’s superior performance highlights its effectiveness in spatial pattern extraction, while GRU offers a computationally efficient alternative. Future work will explore hybrid models and the integration of additional features to enhance prediction robustness. This study contributes to advancing tsunami prediction methodologies, supporting early warning systems for disaster preparedness.