Universitas Ahmad Dahlan Journal
Not a member yet
5744 research outputs found
Sort by
Analisis dan Peningkatan Performa Log File Pada Server dengan Elk Stack
Get PDFKomputer server adalah penyedia layanan dalam jaringan dan memastikan bahwa semua aktivitas komputasi dapat dicatat dan memiliki keluaran berupa file log. Berdasarkan pengamatan yang dilakukan pada sebuah kantor pemerintahan di Kota Padang teridentifikasi kesulitan mengolah file log dan masih memiliki banyak kekurangan dalam proses menganalisa log. Oleh karena itu, dibutuhkan sistem pengelolaan file log secara lengkap, seperti management log yang efektif untuk menampung masalah. Sistem yang diusulkan ini dapat mengevaluasi log jaringan berdasarkan fungsi untuk log management, log collection, log transformation, dan log file. Saat ini telah terdapat banyak alat yang digunakan dengan bermacam metode dalam mengumpulkan log dan menganalisis log untuk mendeteksi aktivitas jahat. Alat yang tersedia belum mengimplementasikan metode ELK Stack, yaitu kombinasi dari Elasticsearch, Logstash, dan Kibana. Arsitektur ELK ini menarik untuk diimplementasikan karena dapat membangun pembaruan yang lebih tertata dengan sedikit usaha dan fungsionalitas yang didapat lebih akurat. Dengan penerapan manajemen log ini, seorang administrator dapat membaca log, meningkatkan performa log file, mengidentifikasi masalah pada server, dan menganalisis pelaporan secara lebih muda
Segmentasi Citra Berdasarkan Kontur Tanah Menggunakan PCA
Get PDFSegmentasi citra berperan penting dalam pemrosesan citra untuk pemetaan tanah, analisis lingkungan, dan aplikasi lainnya. Dalam konteks segmentasi citra berdasarkan kontur tanah, kami mengusulkan pendekatan yang memanfaatkan Analisis Komponen Utama (PCA) untuk memisahkan wilayah tanah dari latar belakang citra. Metode ini bertujuan untuk mengekstraksi fitur-fitur utama dari citra yang mencerminkan variasi intensitas piksel yang signifikan, dengan fokus pada kontur tanah. Langkah-langkah metodologi meliputi pra-pemrosesan citra, seperti normalisasi intensitas, perhitungan komponen utama, seleksi fitur, dan segmentasi berdasarkan kriteria yang ditetapkan. Penggunaan PCA membantu mengatasi masalah dimensi tinggi dalam citra dan memungkinkan ekstraksi informasi yang relevan dengan lebih efisien. Metode ini memiliki potensi aplikasi yang luas dalam pemetaan tanah untuk pertanian presisi, pemantauan lingkungan, dan analisis geologi. Kami mengevaluasi kinerja pendekatan ini menggunakan data citra tanah yang relevan, dan hasil eksperimen menunjukkan keefektifan dan keandalannya dalam memisahkan kontur tanah dari latar belakang citra dengan akurasi yang memuaskan
Meningkatkan enterpreneurial leadership pada mahasiswa melalui experiential learning team project
Get PDFEntreprenurial leardership di Indonesia masih rendah dan perlu diupayakan untuk peningkatannya. Penelitian ini menggunakan metode eksperimen quasi dengan desain one group pretest-posttest design yang memiliki tujuan untuk mengetahui pengaruh metode experential learning team project untuk meningkatkan entrepreneurial leardership pada mahasiswa. Responden penelitian ini sebanyak 32 orang dipilih dengan metode purposive sampling dengan karakteristik mahasiswa psikologi yang belum pernah mengambil mata kuliah psikologi komunitas dan sociopreneurship, serta perkuliahan secara tatap muka di kelas. Hasil penelitian menunjukkan bahwa experential learning team project dapat meningkatkan entrepreneurial leadership mahasiswa (t=-14,15, p<0,05). Artinya, intervensi melalui metode experential learning team project mampu memberikan dampak positif dalam mengembangkan keterampilan enterpreneurial leadership di kalangan mahasiswa psikologi. Implikasi dari penelitian ini adalah bahwa pendekatan experential learning team project dapat menjadi strategi efektif dalam meningkatkan kualitas entrepreneurial leadership di lingkungan pendidikan tinggi.
The influence of social support on students’ academic resilience through self-efficacy
Get PDFStudents who can maintain their abilities and function competently throughout their educational journey until completion are considered resilient. To become a resilient student, several attributes are necessary, both internal and external. Internally, self-efficacy the belief in one’s ability to complete tasks and fulfill obligations is crucial. Externally, social support from family, friends, and significant others plays a vital role. This study aims to examine the effect of social support on students’ academic resilience, with self-efficacy serving as a mediating variable. This study employs a quantitative methodology involving a sample of 270 students from a private university in Medan, with participants aged between 21 and 27 years. The instruments utilized in this research included the MSPSS scale, the General Self-Efficacy Scale, and the Academic Resilience Scale. Mediation analysis conducted using PLS-SEM4 showed that self-efficacy significantly mediates the relationship between social support and academic resilience, as indicated by a t-statistic of 4.440 (greater than 1.96) and a p-value of 0.000 (less than 0.05). This study's findings suggest that self-efficacy partially mediates the relationship between social support and academic resilience among students. Specifically, students who receive strong social support and possess high self-efficacy are more likely to confront their challenges and fulfill their academic responsibilities, ultimately developing resilient students
Design Human Object Detection Yolov4-Tiny Algorithm on ARM Cortex-A72 and A53
Get PDFCurrently, many object detection systems still use devices with large sizes, such as using PCs, as supporting devices, for object detection. This makes these devices challenging to use as a security system in public facilities based on human object detection. In contrast, many Mini PCs currently use ARM processors with high specifications. In this research, to detect human objects will use the Mini PC Nanopi M4V2 device that has a speed in processing with the support of CPU Dual-Core Cortex-A72 (up to 2.0 GHz) + Cortex A53 (Up to 2.0 GHz) and 4 Gb DDR4 Ram. In addition, for the human object detection system, the author uses the You Only Look Once (YOLO) method with the YoloV4-Tiny type, With these specifications and methods, the detection rate and FPS score are seen which are the feasibility values for use in detecting human objects. The simulation for human object recognition was carried out using recorded video, simulation obtained a detection rate of 0,9845 or 98% with FPS score of 3.81-5.55. These results are the best when compared with the YOLOV4 and YOLOV5 models. With these results, it can be applied in various human detection applications and of course robustness testing is needed
Addressing Overfitting in Dermatological Image Analysis with Bayesian Convolutional Neural Network
Get PDFVGG, ResNet, and DenseNet are popular convolutional neural network (CNN) designs for transfer learning (TL), aiding dermatological image processing, particularly in skin cancer categorization. These TL-CNN models build extensive neural network layers for effective image classification. However, their numerous layers can cause overfitting and demand substantial computational resources. The Bayesian CNN (BCNN) technique addresses TL-CNN overfitting by introducing uncertainty in model weights and predictions. Research contributions are (i) comparing BCNN with three TL-CNN architectures in dermatological image processing and (ii) examining BCNN ability to mitigate overfitting through weight perturbation and uncertainty during training. BCNN uses flipout layers to perturb weights during training, guided by the KL divergence and Binary Cross Entropy (BCE) loss function. The dataset used is the ISIC Challenge 2017, categorized as malignant and benign skin tumors. The simulation results show that three TL-CNN architectures, namely VGG-19, ResNet-101, and DenseNet-201, obtained training accuracies of 96.65%, 100%, and 97.70%, respectively. However, all three were only able to achieve a maximum validation accuracy of around 78%. In contrast, BCNN can produce training and validation accuracy of 81.30% and 80%, respectively. The difference in training and validation accuracy values produced by BCNN is only 1.3%. Meanwhile, the three TL-CNN architectures are trapped in an overfitting condition with a difference in training and validation values of around 20%. Therefore, BCNN is more reliable for dermatological image processing, especially for skin cancer images
Genetic Algorithm and GloVe for Information Credibility Detection Using Recurrent Neural Networks on Social Media Twitter (X)
Get PDFSocial media, especially X, has become a key source of information for many individuals, but the level of trust in the information spread on these platforms is a critical issue. To overcome this problem, this research proposed an information credibility detection system using a Recurrent Neural Network (RNN) with the utilization of TF-IDF feature extraction, GloVe feature expansion, BERT word embedding, and Genetic Algorithm (GA) optimization. This research contributes to assessing the credibility of tweets related to the 2024 Indonesian election by integrating TF-IDF to identify important words, GloVe to enhance word context, BERT for deeper understanding, and GA is present to optimize RNN performance. The main focus is to provide maximum accuracy by integrating these methods. In this research, the dataset used consists of 54,766 tweets relating to the 2024 Indonesia election and includes relatively equal numbers of credible and non-credible labels. The corpus construction utilized source X with a total of 40,466 data, IndoNews with a total of 131,580, and a combination of both with a total of 150,943. This research conducted six experimental scenarios, namely optimal data split, max features, N-grams, Top-N rank similarity corpus, BERT and GA application. Through these scenarios, the model achieved a significant accuracy improvement of 1.81% over the baseline, reaching an accuracy of 90.60%. This result demonstrates the effectiveness of the proposed system by presenting a higher quality of accuracy compared to the baseline model. Moreover, this research underscores the significant contribution of increasing the accuracy of information credibility detection
iGWO-RF: an Improved Grey Wolfed Optimization for Random Forest Hyperparameter Optimization to Identification Breast Cancer
Get PDFThe study focuses on improving the accuracy of breast cancer diagnosis by enhancing the predictive capabilities of a Random Forest model. This is achieved by utilizing an improved Grey Wolf Optimization algorithm for hyperparameter optimization. The main objectives are to enhance early detection, increase diagnostic precision, and reduce computational demands in clinical workflows. The work utilizes the Improved Grey Wolf Optimization (iGWO) algorithm to tune the hyperparameters of a Random Forest (RF) model, thereby improving its accuracy in diagnosing breast cancer. The methodology encompasses several steps, including data preparation, model training using iGWO-enhanced RF, performance evaluation compared to traditional methods, and validation using clinical datasets to confirm the reliability and effectiveness of the approach. The iGWO-RF model demonstrated exceptional performance in diagnosing breast cancer, achieving an accuracy of 96.4%, precision of 96.4%, recall of 98.0%, F1-score of 97.2%, and ROC-AUC of 0.988. The findings of iGWO-RF outperform those of SVM, original RF, Naive Bayes, and KNN models, indicating that iGWO-RF is effective in optimizing hyperparameters to improve prediction accuracy. The iGWO-RF model greatly enhances the accuracy and efficiency of breast cancer diagnosis, surpassing conventional models. Integrating iGWO-RF into clinical workflows is advised to improve early identification and patient outcomes. Additional investigation should focus on the utilization of this technology in various medical datasets and circumstances, highlighting its potential in a wide range of healthcare environments
Using Graph Neural Networks and CatBoost for Internet Security Prediction with SMOTE
Get PDFInternet security is the most important issue in cyberspace, on the other hand, cybercrime occurs, and the most serious threat is the theft of personal data and its misuse for the benefit of others. Although cyberspace is while internet security cannot eliminate all risks, predictive models can significantly reduce cybercrime by identifying vulnerabilities if you know how to prevent it. One of the most important things is that many internet users do not know what measures are used to avoid and whether it is safe to visit or explore, on the other hand, in system development existing studies on internet security prediction often rely on generic models that lack precision in identifying influential features or ensuring class balance in developing internet security. In this case, Deep Learning (DL) helps learn patterns from recorded data, find relevant patterns, and use the model effectively. The purpose of this study is to identify the most influential features in internet security and evaluate the effectiveness of advanced machine learning models, such as Graph Neural Networks (GNNs) and Categorical Boosting (CatBoost), for predicting internet safety. So far other studies have tested the entire data set and used a model that is generally. This is expected to lead to the design or development of systems and programs that are useful for internet security. The study used a dataset of 11,055 records with 30 features and binary classification labels ('Safe' and 'Not Safe'). To address the class imbalance, SMOTE was applied before splitting the data into training and testing sets. In testing the Graph Neural Networks (GNNs) model achieved 93.58% accuracy, 93.63% precision, 93.58% recall, and 93.55% F1-score, demonstrating its effectiveness for internet security prediction. From the results of testing the CatBoost model was used to identify key features, revealing that the 'URL of Anchor,' 'SSLFinal State,' and 'Web Traffic' have the most significant impact. From the experiments conducted, the CatBoost effectively identified features with the highest on prediction accuracy, and the GNNs model is very accurate and precise for developing applications or systems to predict internet security
Optimizing 2.4GHz Wireless Networks in Shrimp Ponds with Particle Swarm Optimization
Get PDFThis paper focuses on enhancing wireless sensor networks (WSNs) for monitoring water quality in aquaculture, specifically shrimp ponds, by improving pathloss (PL) models. Radio wave propagation in such environments is challenging due to unpredictable signal attenuation caused by factors like distance, antenna height, terrain, vegetation, and weather conditions. Reliable PL modeling is essential for optimizing network performance. The research evaluates the performance of theoretical PL models, including ITU, Fitting-ITU (FITU), and Weissberger, by comparing their predictions with actual 2.4GHz radio frequency (RF) measurements. Statistical metrics such as root-mean-square error (RMSE) and the coefficient of determination (R²) were used to assess model accuracy. Initial results showed significant discrepancies, with an average RMSE of 28.7dB and an R² of only 5%. To address these issues, the study employed modification techniques (quadratic and cubic polynomial adjustments) and optimization methods, particularly particle swarm optimization (PSO). These approaches refined the theoretical models, aligning them more closely with real-world data. The optimized PSO model reduced the RMSE to 8.34dB and further to 1.89dB, while improving R² from 5% to 95.6%, demonstrating a near-perfect fit. This study highlights the critical role of PSO and similar techniques in bridging the gap between theoretical predictions and practical applications, ensuring more reliable WSN performance in aquaculture environments. The findings contribute to the development of robust, high-accuracy models tailored to the unique challenges of aquaculture settings