Diabetic Nephropathy (DN) is a critical complication with a mortality rate 20-40 times higher than in non-diabetic nephropathy patients, necessitating precise machine learning models to determine whether a patient has nephropathy. Extreme Gradient Boosting (XGBoost) has emerged as a prominent machine learning model for medical diagnostics, with several studies validating its superiority in medical classification. Nevertheless, a significant limitation of XGBoost lies in the complexity of manual hyperparameter tuning. To address this limitation, an automated optimization algorithm is requisite to systematically identify the optimal hyperparameter configuration. This study focuses on optimizing Extreme Gradient Boosting (XGBoost) hyperparameters using Particle Swarm Optimization (PSO), with the F1-Score as its fitness function. To evaluate its effectiveness, the performance of this hybrid XGBoost-PSO model was compared against the baseline XGBoost model. The results showed that the hybrid model outperformed the baseline model, achieving a consistent improvement of 0.02 (2%) across all evaluation metrics. Notably, the F1-Score increased from 0.91 to 0.93, while the Recall metric improved from 0.93 to 0.95. Furthermore, the PSO algorithm efficiently identified the Global Best (GBest) hyperparameters at the 9th iteration. In conclusion, the XGBoost-PSO model provides a robust medical diagnostic tool that maintains a stable performance to enhance clinical judgment.

This study was conducted to predict tidal fluctuations in the coastal area of Gisik Cemandi Village, Sidoarjo, using the Support Vector Regression (SVR) method. The dataset consisted of time series records of sea level height for the period of March . The prediction process was implemented by testing three SVR kernel types, namely linear, polynomial, and Gaussian Radial Basis Function (RBF), along with variations of the parameters Cost , Gamma , and epsilon . Based on the evaluation using Mean Absolute Percentage Error ( MAPE), the Linear kernel demonstrated the best predictive performance with the lowest MAPE value of under a train-test split. The prediction results with the Linear kernel closely matched the actual data, indicating the model’s accuracy and reliability in capturing the linear patterns of tidal data. This model can be utilized as a supporting tool for tidal prediction to aid coastal activities such as navigation and fisheries.

The timeliness of student graduation is an important indicator of academic quality and institutional performance. Delayed graduation not only affects university evaluation metrics but also postpones students’ entry into the workforce. This study proposes a predictive model to identify students at risk of delayed graduation at the Faculty of Computer Science, Universitas Brawijaya. A comparative evaluation of three classification algorithms, namely Random Forest, Support Vector Machine (SVM), and K-Nearest Neighbors (KNN), was conducted within a Knowledge Discovery in Databases (KDD) framework. SMOTE was applied to address class imbalance, while Stratified K-Fold Cross-Validation was used to ensure robust model assessment. Experimental results show that the Random Forest model achieves the best performance, with an accuracy of 73% and an AUC of 0.79, outperforming SVM and KNN. Feature importance analysis further indicates that Grade Point Average, particularly in the third semester, is a more influential predictor of on-time and delayed graduation than credit accumulation. These results demonstrate the potential of the proposed model as an early warning system for proactive academic intervention.

The increasing threat of attacks from quantum computers requires the development of more efficient and secure post-quantum cryptographic algorithms, one of which is NTRU. The main challenge in this algorithm lies in the high complexity of large-dimensional polynomial multiplication operations and parameter sizes that affect system performance. This research implements the Hybridized Number Theoretic Transform and Karatsuba calculation methods with compressed parameters in the C programming language and integrates them into the NTRU algorithm. The evaluation was conducted by measuring the key generation, encryption, and decryption processing times, as well as analyzing the size of the public key, ciphertext, and bandwidth requirements before and after parameter compression. The experimental results show that this method is able to significantly reduce the modulus q value without compromising security, while increasing execution time efficiency. These findings prove that the hybrid NTT–Karatsuba method with compressed parameters supports the practical implementation of the NTRU algorithm in resource-constrained environments.

Low user ratings on mobile applications often reflect underlying dissatisfaction that is not immediately apparent through quantitative scores alone. To uncover the sentiment dynamics behind such evaluations, this study analyzes user reviews of the RCTI+ Superapp in both the original Indonesian and English-translated forms. Using the CRISP-DM framework, reviews were scraped from Google Play, normalized and translated via ChatGPT, and classified using Orange Data Mining with a Naïve Bayes algorithm. The analysis reveals that English-translated reviews yield sharper sentiment polarity and higher classification accuracy (100%) compared to the original Indonesian texts (99.1%), albeit with reduced lexical nuance. These findings suggest that generative AI-assisted translation enhances sentiment clarity in informal, low-resource language data, while potentially simplifying cultural or emotional expression. The study offers methodological insights for multilingual sentiment analysis and practical implications for app developers seeking to interpret user feedback more effectively across languages.

Breast cancer remains one of the leading causes of morbidity and mortality among women worldwide, making early detection essential to improve treatment outcomes. However, early-stage breast cancer symptoms are often subjective and non-specific, which complicates initial risk assessment. This study proposes an expert system for early breast cancer risk diagnosis by integrating Fuzzy Mamdani and Case-Based Reasoning (CBR). The Fuzzy Mamdani method is employed as the primary inference mechanism to model uncertainty in symptoms and risk factors using linguistic rules, while CBR is utilized as a decision support component by leveraging similarities with previously validated clinical cases. The dataset consists of 150 patient records, of which 123 cases are used as the case base and 27 cases are employed for system evaluation. Experimental results show that the proposed system achieves an accuracy of 92.59% compared to expert judgments. These findings indicate that the integration of Fuzzy Mamdani and Case-Based Reasoning provides an interpretable and adaptive approach for early breast cancer risk assessment and has potential as a screening support tool.

This study aims to develop a prediction model for the total Revenue value of the operational activities of the Keputih Surabaya Sewage Sludge Treatment Plant (IPLT) using the Long Short-Term Memory (LSTM) method. The data used is daily data on total transactions and total Revenue from January 2022 to April 2025. Data normalization using the Min-Max method and outlier detection and handling using the IQR and median imputation techniques are examples of preprocessing steps. The model input structure is formed by utilizing Partial Autocorrelation Function (PACF) analysis to ascertain the number of lags. In this study, 405 model combinations are tested with different parameters, including activation function, number of Epochs, learning rate, and ratios of training and testing data. According to the findings, the model that has the optimal parameters a training and testing data ratio of 80:20, 50 Epochs, a learning rate of 0.002, a Tanh activation function, and 100 neurons can produce predictions for total Revenue with a Mean Absolute Percentage Error (MAPE) of 18.18%. The revenue for the following six months was then forecast using this model; the highest revenue forecast was IDR 3,740,085.00, while the lowest was IDR 1,966,628.25. According to these results, LSTM can accurately forecast time series-based income fluctuations and may find use in the waste management industry's financial decision-making and strategic planning processes.

The rapid rise of startups in Indonesia makes user reviews on the Google Play Store a valuable data source for understanding user perceptions and satisfaction. These unstructured reviews contain insights supporting product development and business strategies. This study analyzes sentiments in Indonesian startup app reviews and compares two classification methods: TF-IDF + Linear SVM and fastText, implemented using Google Colab. Reviews were collected in September 2025 using google-play-scraper; 4,000 reviews were retrieved and refined into 3,152 unique reviews after cleaning and preprocessing. Sentiment labeling used ratings (1–2 negative, 4–5 positive); because the neutral class was limited, this study focuses on balanced binary classification with 1576 positive and 1576 negative reviews. The process involves data scraping, text preprocessing, model training, and evaluation using accuracy, precision, recall, and F1-score metrics, with Linear SVM chosen as an efficient baseline for high-dimensional sparse TF-IDF features. Results show that fastText achieves 91.88% accuracy and an F1-macro of 0.9184, slightly outperforming TF-IDF + SVM (F1-macro 0.9103), suggesting that the embedding-based approach better captures semantic nuances of Indonesian text. Future work may extend this study to ABSA to assess sentiments toward price, UI/UX, and customer service for deeper technopreneurship insights in Indonesia.

This study aims to analyse the root causes of the readiness issues faced by the Palu City Government in implementing the concept of an E-Government System. The main focus of this research is to assess the government's readiness in realizing the implementation of E-Government System and to identify the root causes of the obstacles encountered in its application in Palu City. The study uses a qualitative approach, with data analysis conducted using the 3x5 Whys method and a Fishbone diagram. The findings indicate that the implementation of the E-Governemnt System in Palu City faces complex challenges, ranging from unaligned regulations, low human resource competencies, limited infrastructure, to suboptimal funding and an organizational culture that is not yet ready for change. Analysis based on academic theories confirms that the success of the E-Governemnt System implementation does not rely solely on technology but also on institutional readiness and human capacity. This research contributes recommendations that can be utilized by the Palu City Government to improve the quality of the E-Government services and achieve the goals of clean, effective, efficient, transparent, and accountable governance. In addition, this study emphasizes the importance of enhancing human resource capacity and managing change in order to optimize the implementation of the E-Government System in Palu City.