Background:  Foxtail millet (Setaria italica) has great potential as a substitute for wheat flour due to its favorable nutritional composition and functional properties that support the processing of various food products. The utilization of foxtail millet flour in food products is important to promote food diversification, increase the added value of local commodities, and reduce dependence on imported wheat flour. One potential application of foxtail millet flour is as a partial substitute for wheat flour in a traditional cake known as Bay Tat. However, the effect of the percentage of foxtail millet flour substitution on the chemical quality of *bay tat* cake is not yet known.

Aims: This study aimed to determine the effect of different proportions of wheat flour and millet flour on the chemical quality of traditional Bay Tat cake.

Methods: The study employed a Completely Randomized Design (CRD) with a single factor, namely the combination of wheat flour and foxtail millet flour in five treatments: 100% wheat flour and wheat flour substituted with 20%, 30%, 40%, and 50% foxtail millet flour. The observed parameters included moisture content, carbohydrate content, protein content, and dietary fiber content. The data were analyzed statistically using Analysis of Variance (ANOVA), followed by Duncan’s Multiple Range Test (DMRT) for parameters showing significant effects.

Result: The results showed that the substitution of wheat flour with foxtail millet flour significantly affected the chemical quality of traditional Bay Tat cake. Increasing the proportion of foxtail millet flour decreased moisture content (23.12%–16.57%) and carbohydrate content (43.24%–37.97%), while protein content increased (7.88%–14.88%) and crude fiber content increased (0.61%–2.56%).

Conclusion: Millet flour has the potential to be used as a substitute for wheat flour in making Bay Tat to increase protein and fiber content and support food diversification based on local commodities.

Background: Ambon Island is a small mountainous island characterized by high spatial variability of rainfall, which necessitates a representative rainfall gauge network.

Aims: This study aims to evaluate the density and spatial distribution of the rainfall station network on Ambon Island based on the standards of the World Meteorological Organization (WMO) and the Kagan–Rodda method.

Methods: This study uses annual rainfall data from 2014–2023 from eight stations, along with spatial data of the study area. Thiessen polygons were used to define the coverage area of each station in the rainfall station network. The network density was evaluated using WMO standards, and the relationship between distance and rainfall correlation was analyzed using the Kagan–Rodda method to estimate network error.

Result: The results indicate that most stations in the rainfall station network meet WMO density standards, but their spatial distribution is still uneven. The Kagan–Rodda analysis shows that more stations are needed when a smaller error is required. At a 5% error level, three stations are enough with a spacing of about 17.12 km. At a 1% error level, about 51 stations are needed with a spacing of around 4.03 km.

Conclusion: This study highlights the importance of balancing rainfall estimation accuracy with technical and practical limitations in the planning of rainfall gauge networks.

Background: Integrating Artificial Intelligence (AI), ethnoscience, and STEM education is increasingly important for preparing pre-service physics teachers to address technological, interdisciplinary, and culturally responsive demands. However, research in Indonesia remains fragmented, with AI, STEM, and ethnoscience investigated as separate domains, limiting integrated approaches to physics teacher preparation.

Aims: This study identifies publication trends, dominant themes, methodological characteristics, and gaps in integrating AI, ethnoscience, and STEM within Indonesian pre-service physics teacher education.

Methods: Guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework, an integrative literature review examined publications from 2015 to 2026 from Google Scholar, IEEE Xplore, Scopus, and ERIC. Following title, abstract, and full-text screening, 51 articles were selected. Data were analyzed using descriptive bibliometrics, thematic analysis, and gap analysis.

Results: Three trends emerged. First, research emphasizes AI- and STEM-enhanced learning, conceptual understanding, and 21st-century competencies. Second, ethnoscience studies promote local wisdom, contextual learning, cultural literacy, and instructional resources, but remain disconnected from AI-supported STEM learning. Third, comprehensive frameworks integrating AI, ethnoscience, and STEM for pre-service physics teacher education remain scarce. Developmental and qualitative approaches dominate existing studies, while mixed-methods, longitudinal, and intervention-based research remains limited. Gaps persist in examining pre-service teachers’ AI literacy, TPACK, prior knowledge, and capacity to design culturally contextualized STEM learning supported by AI.

Conclusion: The findings reveal a research gap and provide a foundation for developing empirically validated, culturally responsive frameworks integrating AI, STEM, and ethnoscience to strengthen AI literacy, TPACK, STEM competencies, and contextual pedagogical practice in physics teacher education.

Background of study: Diabetes mellitus remains one of the most common chronic diseases worldwide; thus, early and accurate prediction is critical for appropriate intervention.

Aims: This study examines the performance of the standard K-Nearest Neighbor (KNN) method and its fuzzy counterpart, Fuzzy K-Nearest Neighbor (Fuzzy KNN), in classifying diabetes status and compares their behavior across different neighborhood sizes (k).

Methods: The Pima Indians Diabetes Dataset (768 cases, 8 clinical variables) was used. Zero values in physiologically implausible attributes (glucose, blood pressure, skin thickness, insulin, and BMI) were treated as missing values based on the physiological interpretation of these variables and imputed using class-wise medians. The dataset was then split into 80% training and 20% testing subsets using stratified sampling to preserve class proportions. Five-fold cross-validation on the training set was used to determine the optimal number of neighbors (k), after which both models were evaluated on the held-out test set using accuracy, precision, recall, and F1-score.

Result: The optimal k varied between models: k=15 for KNN and k=3 for Fuzzy KNN. Cross-validation accuracy across the tested k range (3–21) remained relatively stable for both methods (approximately 0.80–0.83), indicating that performance was not highly sensitive to k selection. On the test set, Fuzzy KNN outperformed conventional KNN across all evaluation metrics (accuracy: 0.779 vs. 0.773; precision: 0.679 vs. 0.673; recall: 0.704 vs. 0.685; F1-score: 0.691 vs. 0.679).

Conclusion: For the Pima Indians Diabetes Dataset, Fuzzy KNN achieved slightly better performance than conventional KNN across all evaluated metrics on the held-out test set. These findings indicate that distance-based membership weighting may provide a modest advantage over crisp majority voting for diabetes classification in this dataset.