Traditional herbal remedies have an important role in human health. Empirically, Blumea balsamifera is often used as a traditional beverage to alleviate fever symptoms, lower cholesterol levels, and maintain body immunity. The purpose of this study was to discover the phytoconstituent profile that contributes to the anti-diabetic properties of B. balsamifera leaf extract (BBLE) using in silico approaches.LCMS/MS was used to identify the constituent profile of BBLE, and the ability of these compounds against diabetes-related proteins was analyzed computationally.Three proteins related to diabetes are NF-KB p65, GLP-1, and DPP-4. A total of 18 compounds were successfully identified through LCMS/MS, including 4 compounds known to be flavonoid derivatives and can be used as markers of BBLE. Pheophorbide A and 1,1-Cyclopentanediacetic acid were reported for the first time to inhibit the NF-KB p65, GLP-1, and DPP-4 proteins in docking simulation studies. Based on these findings, it can be confirmed that the bioactive compounds in BBLE show strong inhibitory potential against anti-diabetic proteins.
The issue of multimedia users in learning systems still stems from traditional patterns based on paper textbooks and evaluations in the classroom that are vulnerable to cheating, such as copies and leaks of questions from national exams, illegal use of media materials in the form of presentations and websites, and piracy of media substances. This calls for the development of new encryption technologies that incorporate specific data into other digital data, such as learning materials and assessment documents, utilizing the principles of steganography and the techniques of visible watermarking and least significant bit (LSB) encoding. To prevent digital test leaks and cheating during student evaluation, steganography—the science of communicating by using data in digital media so that the data cannot be known by others—is crucial to the evaluation process in this system. This application was developed using the Laravel Framework, Bootstrap V5.0, PHP8.3 JS for DOM software, and a MySQL relational database. The stages of concept, design, material collection, assembly, testing, and distribution are all included in the study process, which employs digital multimedia production using the Luther-Sutopo Method. The outcomes of this application include digital features for instructional materials that can be accessed on the website platform, as well as an interactive learning system.
This research aims to analyze public sentiment and perceptions of the Merdeka Belajar Kampus Merdeka (MBKM) policy through data obtained from social media Twitter. MBKM is a higher education policy in Indonesia that aims to provide tolerance and independence for students in choosing courses and accessing learning resources. The K-Nearest Neighbor (K-NN) method is used to classify sentiment based on text tweets related to MBKM. The steps in this research including tweet data containing keywords related to MBKM taken from Twitter, the text data is extracted, cleaned, and converted into a vector representation, and the K-NN model is drilled using the preprocessed tweet data. This model will classify sentiment into positive, negative, or neutral. Model performance is evaluated using metrics such as accuracy, precision, and recall. It is hoped that the results of this research will provide insight into how society responds to MBKM policies through social media platforms. With a further understanding of these sentiments, educational institutions can take appropriate actions to improve MBKM implementation. The algorithm was applied to 2000 tweet data with the keyword “MBKM”. The model training results prove that the negative precision score is 85%, neutral precision is 53%, positive precision is 70% and accuracy is 71%.
Accessing accurate and comprehensive food calorie and nutrition data remains a challenge, as existing references are often incomplete, outdated, or difficult to access. This limitation reduces the effectiveness of daily dietary monitoring for both individuals and healthcare professionals. This study proposes a method that integrates Bootstrapping Language-Image Pre-training (BLIP) and the Large Language Model (LLaMA) to automatically detect food calories, providing broader coverage of nutritional data. Bootstrapping Language-Image Pre-training generates textual descriptions from food images uploaded by users, which the Large Language Model then processes to estimate key nutritional values, including calories, protein, fat, and carbohydrates. The experimental results demonstrate that this multimodal approach improves both accuracy and efficiency compared to conventional methods. The proposed method offers a practical tool for individuals to independently track daily nutrition and provides healthcare professionals with adaptive AI-based nutritional analysis, contributing to the advancement of digital health technologies in nutrition monitoring and dietary management.