The 2011 flood in Thailand exposed significant vulnerabilities in industrial areas, highlighting the necessity for enhanced disaster risk management through Area-Business Continuity Management (Area-BCM). This preliminary study focuses on identifying how stakeholders rely on information from each other to improve disaster preparedness and response. The research involves systematically identifying Area-BCM stakeholders and designing interview questions, which are evaluated by experts using the Index of Item-Objective Congruence (IOC) to ensure relevance and clarity. All interview questions surpassed the IOC threshold of 0.5, confirming their effectiveness in capturing information interdependencies. Expert feedback led to refinements in the questions, underscoring the importance of tailored data collection. This initial plan provides a critical foundation for understanding information interdependence and highlights the importance of well-designed data collection methods. The findings have significant implications for developing more resilient disaster management strategies in industrial areas, emphasizing the need for precise and relevant stakeholder communication to enhance Area-BCM effectiveness.
In the medical field, the “respiratory rate” of hospitalized patients is measured as one of the vital signs. Respiratory rate is one of the most important items. However, measurement respiratory rate is many burdens of medical personnel because it is done visually. We investigate a method to automatically measure respiratory rate to reduce the burden of medical personnel. An object detection algorithm is applied to the images obtained from the video camera. It is found that it is possible to measure respiratory rate by detecting the width of the subjects bounding box obtained by this method. We propose a method for measuring respiratory rate using this method.
Depression among university students is a critical mental health issue that significantly impacts their academic performance, social interactions, and overall well-being. This study investigates the efficacy of machine learning (ML) and deep learning (DL) models in measuring depression, a significant global mental health concern. Leveraging depression data (PHQ-9) from a survey conducted in the 15 top-ranked universities, our research aims to contribute to students' mental health awareness and support systems with cutting-edge technology. A wide range of ML and DL models are evaluated for this task. Performance is assessed using conventional metrics (e.g., accuracy, precision, and recall), as well as using explainable AI (XAI) to measure the feature extraction capabilities of the models in relation to their prediction accuracy across different dataset sizes. It has revealed that DL models focused more on clinically relevant features in the case of measuring depression. Considering feature importance and performance metrics, different models are applicable for different scenarios, however, overall LSTM, CNN, and SVM are the three top-performing models. Additionally, we found that a hybrid approach, using CNN for feature selection followed by SVM for classification helps to improve the overall accuracy from 0.979 to 0.984. By investigating and bench-marking both ML and DL models, this research establishes a comprehensive basis for the development of effective and easy-to-use AI-based depression measurement tools. Therefore, this study contributes to the early detection and prevention of students' academic depression and supports the well-being of society at large.
This research employed music information retrieval to develop a system for recognizing and classifying Philippine Indigenous Musical Instruments using Custom CNN and VGG-16 architectures. Audio samples were collected from the Katunog database, UPCE archives, and a live recording. The VGG-16 model consistently achieved higher accuracy and F1-scores than the Custom CNN for both individual and family instrument classification, regardless of whether Mel Spectrogram or MFCC data were used. Mel Spectrogram performed better on both CNN architectures. Additionally, most models exhibited higher performance metrics at a k-value of 10 for k-fold cross-validation. The PIMIs were then characterized in terms of timbre and pitch to explore their uniqueness. When using Self-Organizing Maps, MFCC and Mel Spectrogram data showed strong clustering, explaining the high performance of CNN models in recognition and classification tasks.
Recent advancements in artificial intelligence (AI) have led to the development of innovative personalized intelligent assistants such as Google Home and Siri. Despite these technological strides, the integration of established design frameworks and user experience (UX) principles in user-centric AI applications remains insufficiently explored. This study utilizes the design thinking methodology-emphasizing empathy, ideation, and prototype testing-and existing UX laws to develop ‘TimeSync,’ an AI-based smart time management application university students in Thailand. ‘TimeSync’ employs a machine learning model to customize schedules and alarms by incorporating variables such as preparation time, weather, and traffic conditions, ensuring users arrive on time. Usability testing based on specific user satisfaction and usability metrics revealed that ‘TimeSync’ offers an intuitive interface and enhances user experience, promoting consistent utilization for effective time management. These findings demonstrate the significant potential of design thinking and UX principles in crafting user-centric AI applications, providing valuable insights and a robust framework for future AI development.
