Politeknik Negeri Jember (National State of Jember Polytechnic), Indonesia, fully understand that infrastructure is one among various key factors to enhance the quality of the university. However, in the department of Information Technology, infrastructure management is a prominent issue. The primary cause of the issue appears to be a manual arrangement of room reservations, as room usage often crashes due to the scheduling mismatch. To tackle this issue, this study focuses on the development of a room reservation information system, known as “Sistem Informasi Peminjaman Ruangan dan Sarana Pembelajaran (SIPRu)”. The system is developed as an additional feature in the department website. It is developed using the Waterfall method, beginning with the requirements analysis (i.e., an identification of functional and non-functional requirements) and architectural processes design processes (i.e., the development of use case diagram, flowchart, data flow diagram, and database design). In this study, Laravel is used for the implementation, and black box testing is used to verify compliance with the functional requirements. During the testing process, 42 department stakeholders participated to validate the requirements and conduct user experience testing. The results indicated that the stakeholders have found the website to be sufficiently responsive to their needs. Furthermore, it is revealed that SIPRu is effective at assisting users in reserving rooms at the department of Information Technology, National State of Jember Polytechnic.

The process of selecting the most outstanding teachers is an important part of efforts to improve the quality of education in schools. At SMAS Sultan Agung Puger, teacher assessments have so far been subjective and have not used a structured system. Thus, this study aims to design and develop a Decision Support System (DSS) to assist the process of selecting the best teachers based on the evaluation criteria for learning administration that have been set by the school. The method used in this study is the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS), because this method is able to provide objective ranking results oriented towards the level of similarity to the ideal solution. TOPSIS has the advantage of calculating the positive and negative values ​​of each alternative simultaneously. The results of the system implementation show that the teacher with alternative A4 obtained the highest preference value of 1, so he was selected as the best teacher. This system has succeeded in helping to provide more transparent, systematic, and objective evaluation results. Thus, the system built using the TOPSIS method has proven effective in supporting the decision-making process for selecting the best teachers based on learning administration data.

Privacy-preserving messaging for medical device networks

Penelitian ini mengevaluasi pengaruh teknik augmentasi dan fine‑tuning terhadap kinerja model BERT multibahasa pada tugas klasifikasi sentimen ulasan film berbahasa Indonesia. Dataset awal terdiri dari 1.200 ulasan; 80% digunakan untuk pelatihan dan validasi (n = 960) dan 20% untuk pengujian (n = 240). Data pelatihan diperluas melalui augmentasi menjadi 2.880 sampel sintetis untuk keperluan fine‑tuning. Model kemudian di‑fine‑tune pada korpus yang diperluas dan dievaluasi menggunakan metrik akurasi, precision, recall, dan F1. Pada set pengujian diperoleh akurasi 82,5%, precision untuk kelas positif 76,0%, recall 95,0%, dan F1‑score 84,44%. Matriks kebingungan menunjukkan TP = 114, FN = 6, FP = 36, dan TN = 84, yang mengindikasikan sensitivitas tinggi terhadap ulasan positif namun terdapat proporsi false positive yang relatif besar. Temuan ini mengindikasikan bahwa augmentasi meningkatkan kemampuan model dalam menangkap sinyal positif (tingginya recall), namun memerlukan penyesuaian lebih lanjut untuk mengurangi kesalahan prediksi positif (meningkatkan precision). Secara keseluruhan, hasil penelitian menyediakan bukti bahwa BERT multibahasa mampu menangani tugas sentimen berbahasa Indonesia dengan performa memadai apabila didukung strategi augmentasi dan prosedur validasi yang tepat.

Rice leaf disease classification remains challenging due to limited dataset size, class imbalance, and overlapping visual symptoms, which reduce CNN generalization performance. This study investigates whether data augmentation can address these dataset limitations and improve CNN accuracy. The Rice Leaf Diseases Dataset from Kaggle, containing 1,886 images across eight classes, was divided into training, validation, and testing sets (80: 10: 10). Two training schemes were implemented: one with raw images and another with augmentation techniques, including random flipping, rotation, zoom, and contrast adjustment. Models were trained using the Adam optimizer and Sparse Categorical Cross entropy loss. Results showed that augmentation improved classification performance, with validation accuracy rising from 0.51 (non-augmented) to 0.69 (augmented). These findings confirm augmentation as an effective strategy to overcome dataset limitations, enhancing CNN robustness and supporting AI-based agricultural disease diagnosis.

Background: Social-media-based risk screening is promising, yet deployment quality is constrained by noisy language, class imbalance, low explainability, and uncertainty handling. Objective: This work develops an application-specific BiLSTM-centered proxy risk screening framework that combines chaos-regularized learning, SMOTE balancing, SHAP explanation, and TF-IDF similarity fallback. Methods: Experiments used an annotated Twitter corpus (>31,000 English posts) with neutral, hate, and offensive labels, where hate/offensive content was treated only as a proxy risk indicator rather than as evidence of depression, self-harm, or clinical psychological distress. The workflow covered text normalization, tokenization, fixed-length sequence modeling, imbalance correction, and thresholded inference. The model stack used GloVe embeddings with recurrent layers and chaos-based dropout, trained with Adam plus adaptive stopping/scheduling. Operational safety was supported through lexicon checks and a balanced TF-IDF reference set for low-confidence cases. Results: Validation performance reached an F1-score of 0.86, accuracy 0.79, and AUC 0.93 under the reported pipeline. Component-level evidence further indicated that SHAP, thresholding, TF-IDF retrieval, and lexicon checks add explanation, uncertainty handling, reference context, and conservative escalation beyond the underlying classifier, although their marginal effects were not isolated through rerun ablation. Conclusion: The proposed system offers a practical tradeoff between predictive quality, interpretability, and safety-aware operation for proxy risk screening on noisy social media. The contribution is positioned as an integrated application pipeline rather than a new NLP algorithm, and future work should validate the framework with clinically grounded labels, leakage-safe resampling, ablation studies, and transformer baselines.