UIN (Universitas Islam Negeri) Sunan Kalijaga, Yogyakarta: E-Journal Fakultas Sains dan Teknologi
Not a member yet
1440 research outputs found
Sort by
Implementation of Federated Learning for Alzheimer\u27s Disease Classification Using FedAdagrad Algorithm
No full textFederated Learning (FL) offers a promising solution for training machine learning models on decentralized data while preserving privacy, making it particularly valuable for sensitive applications such as healthcare. This study implements FL for the classification of Alzheimer’s disease using MRI images, addressing two critical challenges: data heterogeneity and class imbalance. The research evaluates the performance of the FedAdagrad optimization algorithm against the standard FedAvg approach under varying data distribution scenarios. The methodology employs a CNN trained on a dataset of 6,400 MRI images across four severity classes, partitioned non-IID using Dirichlet distributions (α = 0.1, 0.5, 0.9) to simulate real-world heterogeneity. Experiments were conducted using the Flower framework with four clients over ten communication rounds. Results indicate that FedAdagrad achieves a superior F1-score of 50.33% compared to FedAvg’s 48.14%, though both fall short of centralized CNN performance (55%). High data heterogeneity (α = 0.1) leads to a 13.35% accuracy decline, underscoring FL’s sensitivity to uneven data distributions. Class imbalance emerges as the primary bottleneck, affecting all models. The findings contribute to the growing body of research on adaptive optimization in federated settings, offering insights for future improvements in decentralized healthcare AI
Sentiment Analysis on Shopee Xpress Delivery Time Reviews Using Support Vector Machine and Logistic Regression
No full textThis study examines user sentiment towards Shopee Xpress delivery times using machine learning techniques. We collected 497 reviews from platforms like X and the Google Play Store, leveraging the valuable feedback despite its unstructured and informal nature. After labelling 398 reviews for model training and reserving 99 for sentiment prediction, we implemented two classification algorithms: Support Vector Machine (SVM) and Logistic Regression. These models categorised sentiments into negative, neutral, and positive classes. Despite class imbalance in the training data, SVM outperformed Logistic Regression with an accuracy of 93%, demonstrating a more balanced performance across sentiment categories compared to Logistic Regression\u27s 90% accuracy. Both models showed consistent sentiment prediction on new data. Our findings highlight the potential of sentiment analysis as a valuable tool for Shopee Xpress to understand customer perceptions and improve delivery experiences. By providing actionable insights, this study can inform logistics improvements and enhance customer satisfaction. Future research could benefit from collaborating with Shopee to access internal data and integrating additional data sources for more comprehensive insights, ultimately driving business growth and customer loyalty. This study contributes to the growing body of research on sentiment analysis in logistics and e-commerce
Uncovering Insights in Spotify User Reviews with Optimized Support Vector Machine (SVM)
No full textThe rapid growth of user-generated reviews on platforms like Spotify necessitates efficient analytical techniques to extract valuable insights. This study employs a Support Vector Machine algorithm, optimized using Forward Selection, Backwards Elimination, Optimized Selection, Bagging, and AdaBoost, to effectively classify user reviews. A dataset of approximately 10,000 Spotify reviews was compiled from diverse online sources, ensuring a representative sample. The analysis reveals sentiment patterns across positive, negative, and neutral categories, with positive reviews dominates the landscape. These patterns help highlight Spotify’s strengths while identifying areas for improvement. However, the SVM algorithm faces challenges in classifying minority classes, particularly negative sentiments, due to class imbalance. To address this, advanced optimization techniques are utilized to enhance classification precision and recall. Preprocessing steps, including data cleansing, tokenization, stemming, and stopword removal, refine the dataset, while TF-IDF converts text into numerical features for effective feature selection. The results show that the Optimized Selection method achieves the highest accuracy of 84.5%, outperforming other approaches. This research contributes significantly to developing balanced sentiment analysis models. Future studies may explore deep learning techniques to further improve classification accuracy and mitigate current limitations in data representation
Kajian Islam Tentang Sifat Boros dan Implementasinya Pada Pembuatan Sistem Pemantauan serta Pengendalian Bahan Baku di Gudang
No full textPenelitian ini bertujuan untuk menerapkan wawasan Islam tentang menghindari pemborosan dalam merancang sistem pemantauan dan pengendalian gudang. Islam menekankan pentingnya menghindari perilaku pemborosan, khususnya dalam penggunaan sumber daya seperti bahan baku, energi, tenaga kerja, dan waktu. Penelitian ini dilakukan di PT. Citra Sarungtangan Indonesia yang menghadapi masalah kerusakan bahan baku akibat kondisi penyimpanan yang tidak tepat. Penelitian ini menggunakan pendekatan mixed-method, termasuk tinjauan pustaka dan pengembangan sistem berbasis Internet of Things (IoT) yang memanfaatkan sensor DHT22 dan NodeMCU ESP8266. Hasilnya adalah sistem otomatis yang memantau dan mengontrol suhu dan kelembaban gudang, yang dapat diakses melalui aplikasi IoT MQTT Panel. Implementasi sistem ini berhasil meningkatkan efisiensi penyimpanan dan mengurangi pemborosan bahan baku, sejalan dengan prinsip-prinsip Islam tentang menghindari pemborosan
Imputasi Data Temperatur Maksimum Menggunakan Metode Support Vector Regression
No full textTemperature is a crucial element affecting various aspects, from agriculture to natural disasters. Temperature data imputation is also important because, in some cases, temperature data is not always complete. This study aims to predict missing temperature data in the East Nusa Tenggara (NTT) region using the Support Vector Regression (SVR) method. The data used comes from six BMKG observation stations in NTT and ERA-5 Reanalysis data. The choice of the SVR method is based on its ability to handle data with complex structures. Modeling is conducted separately for each station using the Radial Basis Function (RBF) kernel. Model evaluation employs the metrics Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Coefficient of Determination (R²), presenting the evaluation results with low error. The results show that among several parameter tests, the parameter ranges [C = 1, 5, 10, 15], [ε = 0,1, 0,3, 0,6, 0,9], and [γ = 1, 5, 10, 15] in the SVR method are the best parameter ranges across all stations. The prediction graphs display different temperature fluctuation patterns at each station. This study contributes to enhancing the availability of accurate climate data, supporting sustainable decision-making in the NTT region
Penggunaan Teknik Transfer Learning pada Metode CNN untuk Pengenalan Tanaman Bunga
No full textThis study investigates the impact of employing the transfer learning method on improving flower recognition performance using Convolutional Neural Network (CNN) models. The dataset used consists of 4242 flower images divided into five classes: daisy, tulip, rose, sunflower, and dandelion. This research implements three models: basic CNN, VGG16, and EfficientNetB3, to test the effectiveness of transfer learning in flower classification. The basic CNN model achieved a training accuracy of 73.38% and a validation accuracy of 71.76%, but it generally fails to generalize to new data. The VGG16 model achieved perfect training accuracy but experienced overfitting, with validation accuracy stabilizing around 85-90%. Meanwhile, the EfficientNetB3 model with transfer learning reached a training accuracy of 98.50% and a validation accuracy of 94.00%, demonstrating strong generalization without significant overfitting. The experiment was conducted using data augmentation techniques, and performance evaluation was carried out using accuracy, precision, and recall metrics. The results show that transfer learning with the EfficientNetB3 model provides the best performance in flower classification compared to the basic CNN and VGG16 models. For future research, further development can be done by expanding the types of flower datasets and applying additional optimization techniques to improve accuracy in more complex models
Perbandingan Sensitivitas Metode SAW, MAUT dan WSM pada Anugerah Mutu Non-Akademik Universitas
No full textA Decision Support System works best with a suitable method. Unfortunately, not all methods are equally used. Two rarely used methods are the MAUT and WSM methods. To determine whether a method is more suitable for a case than another, a sensitivity test is conducted. By conducting sensitivity tests between the two methods and other commonly used methods, such as SAW, in the same case, it’s possible to compare the sensitivity percentages of the three. One case that can be helped by a Decision Support System using the three methods is the ANOMIK assessment at universities. The three methods produced the same best alternative, namely Faculty 9. After conducting a sensitivity test, the results showed that the WSM method was the most sensitive, with a value of 4.954%, followed by the SAW method with a value of 4.901%, and finally the MAUT method with a value of 3.844%
Application of SMOTE in Sentiment Analysis of MyXL User Reviews on Google Play Store
No full textTexts that express customer opinions about a product are important input for companies. Companies obtain valuable information from consumer perceptions of marketed products by conducting sentiment analysis. However, real-world text datasets are often unbalanced, causing the prediction results of classification algorithms to be biased towards the majority class and ignoring the minority class. This study analyzes the sentiment of MyXL user reviews on the Google Play Store by comparing the performance of the Logistic Regression and Support Vector Machine algorithms in the SMOTE implementation. This analysis uses TF-IDF to extract features and GridSearchCV to optimize the accuracy, precision, recall, and F1-score evaluation metrics. This study follows several scenarios of dividing training data and test data. SVM implementing SMOTE is the algorithm with the best performance using the division of training data (90%) and test data (10%), resulting in accuracy (73.00%), precision (67.13%), recall (65.82%), and F1-score (66.30%)
Analisis Pengaruh Kompresi File Pada Media Sosial Terhadap Ketahanan Image Steganografi Pada Metode Least Significant Bit (LSB)
No full textMedia sosial saat ini menjadi platform utama untuk pertukaran informasi di kalangan masyarakat luas. Platform seperti Facebook, Instagram, dan Twitter memungkinkan pengguna untuk berbagi gambar dengan audiens yang sangat besar. Namun, penggunaan media sosial ini juga menimbulkan kekhawatiran terkait privasi dan keamanan data, karena informasi yang dibagikan di platform ini rentan terhadap ancaman kejahatan. Metode Least Significant Bit (LSB) merupakan salah satu teknik steganografi yang paling sederhana dan paling banyak digunakan dalam penyembunyian data. Tujuan dari penelitian ini adalah untuk mengembangkan dan mengimplementasikan metode steganografi LSB yang lebih aman dan tahan terhadap serangan dan gangguan. Metode Least Significant Bit (LSB) merupakan salah satu teknik steganografi yang efektif untuk menyembunyikan pesan dalam media digital, termasuk gambar. Namun efektivitas metode ini dapat dianalisis berdasarkan beberapa parameter, seperti kapasitas penyimpanan, keterlihatan, dan ketahanan terhadap kompresi atau manipulasi gambar, khususnya pada platform media sosial seperti Telegram, Instagram, dan Facebook. Hasil pengujian yang dilakukan menunjukkan, file pesan yang telah disembunyikan dalam gambar tidak dapat ditemukan lagi akibat perubahan ekstensi dari file gambar yang telah di kirim serta penerapan metode kompresi lossy pada masing-masing platform media sosial juga mempengaruhi file stego yang ada.
Kata kunci: Digital Forensik, Sosial Media, Gambar, Steganografi, Least Significant Bit
--------------------------------------------------------------------------------------------------
Analysis Of The Effect Of File Compression On Social Media On Image Steganography Resilience In The Least Significant Bit (LSB) Method
Social media is currently a major platform for information exchange among the wider community. Platforms such as Facebook, Instagram, and Twitter allow users to share images with a very large audience. However, the use of social media also raises concerns regarding privacy and data security, because information shared on these platforms is vulnerable to threats of crime. The Least Significant Bit (LSB) method is one of the simplest and most widely used steganography techniques in data hiding. The purpose of this study is to develop and implement a LSB steganography method that is more secure and resistant to attacks and interference. The Least Significant Bit (LSB) method is one of the effective steganography techniques for hiding messages in digital media, including images. However, the effectiveness of this method can be analyzed based on several parameters, such as storage capacity, visibility, and resistance to compression or image manipulation, especially on social media platforms such as Telegram, Instagram, and Facebook. Test results show that message files that have been hidden in images can no longer be found due to changes in the extension of the image file that has been sent. The application of lossy compression methods on each social media platform also affects the existing stego file..
Keywords: Digital Forensics, Social Media, Image, Steganography, Least Significant Bi
Subsurface Structure Analysis for Determining the Slip Surface of Landslides Using the Wenner Resistivity Geoelectrical Method in Kokap, Kulon Progo: Analisis Struktur Bawah Permukaan untuk Penentuan Bidang Geser Longsor Menggunakan Metode Geolistrik Resistivitas Wenner di Kokap, Kulon Progo
No full textSlope failures frequently occur in hilly regions, particularly during periods of intense rainfall. At the end of 2022, such an event affected a residential area in Hargomulyo Village, Kulon Progo. Mitigation efforts against similar hazards can be implemented through mapping of landslide-prone zones, one of which involves identifying the rock layers that act as the slip surface. This study employs the geoelectrical resistivity method with a Wenner configuration to characterize the subsurface structure based on variations in electrical resistivity of soil and rock. Data acquisition was conducted along four survey lines located within an andesitic intrusion formation composed of hypersthene–andesite to trachyandesite rocks. The modeling results indicate three main layers: surface soil with resistivity values below 54.4 Ωm, clay mixed with weathered andesite ranging from 54.4 Ωm to 141 Ωm, and intact andesite bedrock exceeding 141 Ωm. Correlation with the Geological Map of Yogyakarta Sheet (Rahardjo et al., 1995) confirms that the potential slip surface corresponds to the weathered andesite–clay zone developed along the contact between the Andesite Intrusion and the Kebobutak Formation. This transition layer, acting as a mechanically weak zone, controls slope stability in the study area. The presence of a local north–south fault may further increase groundwater infiltration and pore pressure above the impermeable andesite layer, promoting slope movement during heavy rainfall. These results emphasize that integrating resistivity interpretation with geological mapping provides a more comprehensive understanding of landslide mechanisms and supports hazard mitigation efforts in andesitic terrains of Kulon Progo