Claim Missing Document
Check
Articles

Comparative Analysis of Neural Network Architectures for Mental Health Diagnosis: A Deep Learning Approach Airlangga, Gregorius
KLIK: Kajian Ilmiah Informatika dan Komputer Vol. 4 No. 4 (2024): Februari 2024
Publisher : STMIK Budi Darma

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30865/klik.v4i4.1703

Abstract

Mental health conditions present a complex diagnostic challenge due to the subtlety and diversity of symptoms. This study provides a comprehensive analysis of various neural network architectures, including Multilayer Perceptron (MLP), Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Long Short-Term Memory networks (LSTM), and Dense Neural Network (DNN), in their ability to classify mental health conditions. Utilizing a rich dataset of symptoms and expert diagnoses, we preprocessed the data to address class imbalances and trained each model to evaluate its diagnostic performance. Our results are presented through confusion matrices that reveal the accuracy, precision, recall, and F1-scores for each model. The MLP and DNN models demonstrated high accuracy in identifying distinct conditions but struggled with overlapping symptoms. LSTM and RNN models captured temporal patterns to some extent yet required further optimization for improved accuracy. CNN models showed robust feature detection capabilities, with the CNN 1D model excelling in specificity for certain conditions. However, a common challenge across all models was the differentiation between conditions with similar symptom presentations. Our findings suggest that while individual models have their strengths, an ensemble approach may be necessary for enhanced diagnostic precision. Future work will focus on integrating models, refining feature extraction, and employing explainable AI to increase transparency and trust in model predictions. Additionally, expanding the dataset and conducting clinical trials will ensure the models' effectiveness in real-world settings. This research moves us closer to achieving nuanced, AI-driven diagnostics that can support clinicians and benefit patient outcomes in mental healthcare.
A Comparative Analysis of Clustering Algorithms for Expedia’s Travel Dataset Airlangga, Gregorius
Sinkron : jurnal dan penelitian teknik informatika Vol. 9 No. 1 (2025): Research Article, January 2025
Publisher : Politeknik Ganesha Medan

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33395/sinkron.v9i1.14343

Abstract

The effective segmentation of travel data is crucial for deriving actionable insights in the tourism and hospitality sectors. This study conducts a comprehensive evaluation of four clustering algorithms Agglomerative Clustering, DBSCAN, Gaussian Mixture Models (GMM), and KMeans on a travel dataset, using three widely recognized metrics: Silhouette Score, Davies-Bouldin Index, and Calinski-Harabasz Score. The dataset was preprocessed through standardization and dimensionality reduction via Principal Component Analysis (PCA) to facilitate visualization and ensure computational efficiency. The results highlight significant differences in the performance of these algorithms. Agglomerative Clustering achieved the highest Silhouette Score, indicating superior cluster cohesion and separation, while KMeans recorded the highest Calinski-Harabasz Score, demonstrating strong inter-cluster variance. In contrast, DBSCAN performed poorly, producing low scores across all metrics, attributed to sensitivity to parameter selection and density irregularities in the dataset. Gaussian Mixture Models exhibited moderate performance but struggled with overlapping clusters due to limitations in modeling non-Gaussian data distributions. Visualization of clustering results confirmed these findings, revealing compact clusters for Agglomerative and KMeans, while DBSCAN and GMM showed less defined structures. This study underscores the importance of selecting clustering algorithms based on dataset characteristics and analysis objectives
A Comparative Study of Ensemble Learning and Neural Networks for the Heart Disease Prediction Airlangga, Gregorius; Nugroho, Oskar Ika Adi; Lim, Bobi Hartanto Pramudita
Sinkron : jurnal dan penelitian teknik informatika Vol. 9 No. 1 (2025): Research Article, January 2025
Publisher : Politeknik Ganesha Medan

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33395/sinkron.v9i1.14347

Abstract

Heart disease continues to be a leading global cause of death, making the development of predictive models for early diagnosis a critical task. This study investigates the performance of various machine learning and deep learning models for heart disease prediction using a structured dataset of 918 observations and 11 features. The analysis includes ensemble methods like Random Forest, Gradient Boosting, and XGBoost, as well as neural networks such as Multi-Layer Perceptrons (MLPs) and Convolutional Neural Networks (CNNs). Traditional classifiers, including Support Vector Machines (SVM) and Logistic Regression, are also considered for benchmarking. The dataset was preprocessed using label encoding, standardization, and the Synthetic Minority Oversampling Technique (SMOTE) to address class imbalance and ensure data consistency. Model evaluation was conducted using key metrics such as precision, recall, F1-score, and ROC-AUC. The results demonstrated that ensemble methods, particularly Random Forest (ROC-AUC: 0.9313) and Gradient Boosting (ROC-AUC: 0.9279), consistently delivered superior performance. Among neural networks, MLPs showed promising results (ROC-AUC: 0.9232), outperforming CNNs, which were less effective in handling tabular data. Meanwhile, TabNet was found to be unsuitable for this dataset, as it significantly underperformed across all metrics. This research highlights the effectiveness of ensemble methods and MLPs in heart disease prediction and the importance of proper preprocessing techniques. Future work could focus on integrating hybrid models or advanced optimization techniques to further enhance predictive accuracy in clinical settings.
Application of Traditional Machine Learning Techniques for the Classification of Human DNA Sequences: A Comparative Study of Random Forest and XGBoost Airlangga, Gregorius
Jurnal Informatika Universitas Pamulang Vol 9 No 1 (2024): JURNAL INFORMATIKA UNIVERSITAS PAMULANG
Publisher : Teknik Informatika Universitas Pamulang

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.32493/informatika.v9i1.39353

Abstract

This study evaluates the performance of hybrid machine learning models, specifically Random Forest and XGBoost, in classifying human DNA sequences into seven functional classes. Utilizing advanced feature vectorization techniques, this research addresses the challenges of analyzing high-dimensional genomic data. Both models were trained and tested on a dataset of annotated human DNA sequences, with an emphasis on generalizability to new, unseen data. Our results indicate that the Random Forest model achieved an accuracy of 87.98%, slightly outperforming the XGBoost model, which recorded an accuracy of 87.06%. These findings underscore the effectiveness of employing traditional machine learning techniques coupled with innovative data preprocessing for predictive modeling in genomics. The study not only enhances our understanding of genomic functionalities but also suggests robust methodologies for future genetic research and potential applications in personalized medicine. The implications of these results for improving classification accuracy and the recommendations for integrating more complex algorithms are also discussed
A Hybrid Model for Human DNA Sequence Classification Using Convolutional Neural Networks and Random Forests Airlangga, Gregorius
Jurnal Informatika Universitas Pamulang Vol 9 No 2 (2024): JURNAL INFORMATIKA UNIVERSITAS PAMULANG
Publisher : Teknik Informatika Universitas Pamulang

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.32493/informatika.v9i2.39355

Abstract

Human DNA sequence classification is a fundamental task in genomics, essential for understanding genetic variations and its implications in disease susceptibility, personalized medicine, and evolutionary biology. This study proposes a novel hybrid model combining Convolutional Neural Networks (CNN) for feature extraction and Random Forest classifiers for final classification. The model was evaluated on a dataset of human DNA sequences, with achieving an accuracy of 75.34%. The results showed that performance metrics, including precision, recall, and F1-scores across multiple classes, showed significant improvements over traditional models. The CNN component effectively captures local dependencies and patterns within the sequences, while the Random Forest classifier handles complex decision boundaries, resulting in enhanced classification accuracy. Comparative analysis demonstrated the superiority of our hybrid approach, with the CNN-LSTM model achieving only 59.47% accuracy, and other RNN-based models like CNN-GRU and CNN-BiLSTM performing similarly lower. These results suggest that hybrid models can leverage the strengths of both deep learning and traditional machine learning techniques an offering a more effective tool for DNA sequence classification. The future work will optimize model architecture and explore larger, thus more diverse datasets to validate our approach's generalizability and robustness.
Fuzzy A* for optimum Path Planning in a Large Maze Airlangga, Gregorius
Buletin Ilmiah Sarjana Teknik Elektro Vol. 5 No. 4 (2023): December
Publisher : Universitas Ahmad Dahlan

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.12928/biste.v5i4.9394

Abstract

 Traditional A* path planning, while guaranteeing the shortest path with an admissible heuristic, often employs conservative heuristic functions that neglect potential obstacles and map inaccuracies. This can lead to inefficient searches and increased memory usage in complex environments. To address this, machine learning methods have been explored to predict cost functions, reducing memory load while maintaining optimal solutions. However, these require extensive data collection and struggle in novel, intricate environments. We propose the Fuzzy A* algorithm, an enhancement of the classic A* method, incorporating a new determinant variable to adjust heuristic cost calculations. This adjustment modulates the scope of scanned vertices during searches, optimizing memory usage and computational efficiency. In our approach, unlike traditional A* heuristics that overlook environmental complexities, the Fuzzy A* employs a dynamic heuristic function. This function, leveraging fuzzy logic principles, adapts to varying levels of environmental complexity, allowing a more nuanced estimation of the path cost that considers potential obstructions and route feasibility. This adaptability contrasts with standard machine learning-based solutions, which, while effective in known environments, often falter in unfamiliar or highly complex settings due to their reliance on pre-existing datasets. Our experimental framework involved 100 maze-solving trials in diverse maze configurations, ranging from simple to highly intricate layouts, to evaluate the effectiveness of Fuzzy A*. We employed specific metrics such as path length, computational time, and memory usage for a comprehensive assessment. The results showcased that Fuzzy A* consistently found the shortest paths (99.96% success rate) and significantly reduced memory usage by 67% and 59% compared to Breadth-First-Search (BFS) and traditional A*, respectively. These findings underline the effectiveness of our modified heuristic approach in diverse and challenging environments, highlighting its potential for real-world pathfinding applications.
Advancing UAV Path Planning System: A Software Pattern Language for Dynamic Environments Airlangga, Gregorius
Buletin Ilmiah Sarjana Teknik Elektro Vol. 5 No. 4 (2023): December
Publisher : Universitas Ahmad Dahlan

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.12928/biste.v5i4.9407

Abstract

In the rapidly advancing domain of Unmanned Aerial Vehicle (UAV) technologies, the capability to navigate dynamic and unpredictable environments is paramount. To this end, we present a novel design pattern framework for real-time UAV path planning, derived from the established Pattern Language of Program Community (PLOP). This framework integrates a suite of software patterns, each selected for its role in enhancing UAV operational adaptability, environmental awareness, and resource management. Our proposed framework capitalizes on a blend of behavioral, structural, and creational patterns, which work in concert to refine the UAV's decision-making processes in response to changing environmental conditions. For instance, the Observer pattern is employed to maintain real-time environmental awareness, while the Strategy pattern allows for dynamic adaptability in the UAV's path planning algorithm. Theoretical analysis and conceptual evaluations form the backbone of this research, eschewing empirical experiments for a detailed exploration of the design's potential. By offering a systematic and standardized approach, this research contributes to the UAV field by providing a robust theoretical foundation for future empirical studies and practical implementations, aiming to elevate the efficiency and safety of UAV operations in dynamic environments.
Adaptive Cyber-Defense for Unmanned Aerial Vehicles: A Modular Simulation Model with Dynamic Performance Management Airlangga, Gregorius
Buletin Ilmiah Sarjana Teknik Elektro Vol. 5 No. 4 (2023): December
Publisher : Universitas Ahmad Dahlan

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.12928/biste.v5i4.9415

Abstract

In light of escalating cyber threats, this study tackles the cybersecurity challenges in UAV systems, underscoring the limitations of static defense mechanisms. Traditional security approaches fall short against the sophisticated and evolving nature of cyber-attacks, particularly for UAVs that depend on real-time autonomy. Addressing this deficiency, we introduce an adaptive modular security system tailored for UAVs, enhancing resilience through real-time defensive adaptability. This system integrates scalable, modular components and employs machine learning techniques—specifically, neural networks and anomaly detection algorithm to improve threat prediction and response. Our approach marks a significant leap in UAV cybersecurity, departing from static defenses to a dynamic, context-aware strategy. By employing this system, UAV stakeholders gain the flexibility needed to counteract multifaceted cyber risks in diverse operational scenarios. The paper delves into the system's design and operational efficacy, juxtaposing it with conventional strategies. Experimental evaluations, using varied UAV scenarios, measure defense success rates, computational efficiency, and resource utilization. Findings reveal that our system surpasses traditional models in defense success and computational speed, albeit with a slight increase in resource usage a consideration for deployment in resource-constrained contexts. In closing, this research underscores the imperative for dynamic, adaptable cybersecurity solutions in UAV operations, presenting an innovative and proactive defense framework. It not only illustrates the immediate benefits of such adaptive systems but also paves the way for ongoing enhancements in UAV cyber defense mechanisms.
Comparative Analysis of MLP, CNN, and RNN Models in Automatic Speech Recognition: Dissecting Performance Metric Lenson, Abraham K. S.; Airlangga, Gregorius
Buletin Ilmiah Sarjana Teknik Elektro Vol. 5 No. 4 (2023): December
Publisher : Universitas Ahmad Dahlan

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.12928/biste.v5i4.9668

Abstract

This study conducts a comparative analysis of three prominent machine learning models: Multi-Layer Perceptrons (MLP), Convolutional Neural Networks (CNN), and Recurrent Neural Networks (RNN) with Long Short-Term Memory (LSTM) in the field of automatic speech recognition (ASR). This research is distinct in its use of the LibriSpeech 'test-clean' dataset, selected for its diversity in speaker accents and varied recording conditions, establishing it as a robust benchmark for ASR performance evaluation. Our approach involved preprocessing the audio data to ensure consistency and extracting Mel-Frequency Cepstral Coefficients (MFCCs) as the primary features, crucial for capturing the nuances of human speech. The models were meticulously configured with specific architectural details and hyperparameters. The MLP and CNN models were designed to maximize their pattern recognition capabilities, while the RNN (LSTM) was optimized for processing temporal data. To assess their performance, we employed metrics such as precision, recall, and F1-score. The MLP and CNN models demonstrated exceptional accuracy, with scores of 0.98 across these metrics, indicating their effectiveness in feature extraction and pattern recognition. In contrast, the LSTM variant of RNN showed lower efficacy, with scores below 0.60, highlighting the challenges in handling sequential speech data. The results of this study shed light on the differing capabilities of these models in ASR. While the high accuracy of MLP and CNN suggests potential overfitting, the underperformance of LSTM underscores the necessity for further refinement in sequential data processing. This research contributes to the understanding of various machine learning approaches in ASR and paves the way for future investigations. We propose exploring hybrid model architectures and enhancing feature extraction methods to develop more sophisticated, real-world ASR systems. Additionally, our findings underscore the importance of considering model-specific strengths and limitations in ASR applications, guiding the direction of future research in this rapidly evolving field.
Optimizing UAV Navigation: A Particle Swarm Optimization Approach for Path Planning in 3D Environments Airlangga, Gregorius
Buletin Ilmiah Sarjana Teknik Elektro Vol. 5 No. 4 (2023): December
Publisher : Universitas Ahmad Dahlan

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.12928/biste.v5i4.9696

Abstract

This study explores the application of Particle Swarm Optimization (PSO) in Unmanned Aerial Vehicle (UAV) path planning within a simulated three-dimensional environment. UAVs, increasingly prevalent across various sectors, demand efficient navigation solutions that account for dynamic and unpredictable elements. Traditional pathfinding algorithms often fall short in complex scenarios, hence the shift towards PSO, a bio-inspired algorithm recognized for its adaptability and robustness. We developed a Python-based framework to simulate the UAV path planning scenario. The PSO algorithm was tasked to navigate a UAV from a starting point to a predetermined destination while avoiding spherical obstacles. The environment was set within a 3D grid with a series of waypoints, marking the UAV's trajectory, generated by the PSO to ensure obstacle avoidance and path optimization. The PSO parameters were meticulously tuned to balance the exploration and exploitation of the search space, with an emphasis on computational efficiency. A cost function penalizing proximity to obstacles guided the PSO in real-time decision-making, resulting in a collision-free and optimized path. The UAV's trajectory was visualized in both 2D and 3D perspectives, with the analysis focusing on the path's smoothness, length, and adherence to spatial constraints. The results affirm the PSO's effectiveness in UAV path planning, successfully avoiding obstacles and minimizing path length. The findings highlight PSO's potential for practical UAV applications, emphasizing the importance of parameter optimization. This research contributes to the advancement of autonomous UAV navigation, indicating PSO as a viable solution for real-world path planning challenges.