Machine learning is used for a fast pre-diagnosis approach to prevent the effects of Major Depressive Disorder (MDD). The objective of this research is to detect depression using a set of important ...facial features extracted from interview video, e.g., radians, gaze at angles, action unit intensity, etc. The model is based on LSTM with an attention mechanism. It aims to combine those features using the intermediate fusion approach. The label smoothing was presented to further improve the model's performance. Unlike other black-box models, the integrated gradient was presented as the model explanation to show important features of each patient. The experiment was conducted on 474 video samples collected at Chulalongkorn University. The data set was divided into 134 depressed and 340 non-depressed categories. The results showed that our model is the winner, with a 88.89% F1-score, 87.03% recall, 91.67% accuracy, and 91.40% precision. Moreover, the model can capture important features of depression, including head turning, no specific gaze, slow eye movement, no smiles, frowning, grumbling, and scowling, which express a lack of concentration, social disinterest, and negative feelings that are consistent with the assumptions in the depressive theories.
The third-generation meteorological satellites equipped with highly-improved imagers provide a huge amount of Earth observation data. Himawari-8 is the first unit of the Japan Meteorological Agency’s ...third generation of geostationary satellites. After its starting operation in 2015, there are several websites that provide remotely sensed images in real time. However, it is hard to develop a real-time and full-resolution website, due to the large amount of data to be handled. Himawari-8 real-time web is only one website that provides full-resolution remotely sensed images in real time. To reduce network traffic and increase the access speed of it from other countries out of Japan, mirror websites of each country are needed. In this paper, we propose a cost-effective mirroring system for the Himawari-8 real-time web. A mirroring model is introduced to avoid the problem of big data processing in the mirror websites. We adopt a file copy tool based on high-performance and flexible protocol (HpFP) to transfer meteorological satellite data from the Himawari-8 real-time web to the mirror websites. Our first target is Thailand, one of the most disaster-prone countries in South-East Asia. The mirror website is set up at an institution in Thailand connected via collaborative international networks, e.g., Japan Gigabit Network (JGN) and Asia Pacific Advanced Network (APAN). The results show that the proposed mirroring system is able to overcome the big data issue by reducing the central processing unit (CPU) usage in the mirror website and transferring remotely sensed image files at high speed over international networks even under packet loss conditions. This suggests that our mirroring system has a potential for deployment in other Asian and Oceanian countries.
The national digital identity platform is a robust scheme that enables individuals and entities to prove who they are to digitally access critical information or services. However, c urrent digital ...identity systems do not sufficiently consider delegation between entities from the viewpoint of dynamic authorizers and permissions. This study aims to understand the pain points and expectations of end-users and service providers in the Thai national digital identity platform, to design a dynamic delegation model and develop an innovative delegation application to test user acceptance. The research utilizes semi-structured interviews with 3 digital identity experts, two focus groups, one with 6 service providers, and the other one with 6 end-users. Based on results from the data analysis and conceptual prototype design, validated by experts, the proposed prototype is practical and suitable for developing a digital delegation mobile web application that is convenient, safe, secure, and reliable utilizing blockchain technology under the Thai national digital identity platform. The technology acceptance model was used to test the application acceptance with 42 participants . The result reveals that both person and businesses intend to adopt the digital delegation mobile web application. Use cases of the application include users give their power to trusted entities and Government Agency to provide services to the citizens via the authorized delegatee.
Electronic cognitive assessment tools are essential in screening and managing patients’ disease progression in clinical practices. Neurocognitive disorders, particularly mild cognitive impairment ...(MCI) and Alzheimer’s Dementia (AD), can be assessed using the Montreal Cognitive Assessment (MoCA) 1, 2. A digital version of the MoCA could help reduce the burden on healthcare providers, increase accessibility, and maximize its utility, especially in countries that lack medical staff such as Thailand. This pilot study aimed to establish the concurrent validity between the electronic version of the MoCA‐Thai (eMoCA‐Thai), developed by the authors of this paper, to the original paper‐and‐pencil version (MoCA‐Thai) in adults who are cognitively normal (CN), have MCI, and have AD. 54 adults between the ages of 60 to 90 attending a dementia clinic were administered both the eMoCA‐Thai and the standard MoCA‐Thai one week apart. The primary outcome measures include the total scores and all subscale scores of the two test versions. Correlations and differences between scores were analyzed using concordance correlation coefficients (CCC) and various parametric t‐tests. Results show that there is high correlation for total MoCA scores with a CCC of 0.919 and a mean difference of ‐0.204 (‐6.311, 5.904). All the cognitive subdomain scores had moderate to high CCC of more than 0.4. The differences in average total score (22.81±7.49 for the MoCA‐Thai and 23.02±7.99 for the eMoCA‐Thai) were not statistically significant (p = 0.633), indicating no differences between the MoCA scores of both versions. When comparing total MoCA scores between participant groups, both CN and MCI had acceptable correlations (CCC >0.2) with a mean difference of 0.355 (‐4.720, 5.429) and ‐1.643 (9.001, 5.715), respectively, whilst AD had excellent correlation (CC >0.8) with a mean difference of 0.111 (‐6.496, 6.717). Subdomain scores between participant groups were not statistically significantly different except for Delayed Recall for all participants combined (p<0.05). In conclusion, there is adequate concurrent validity between the MoCA‐Thai and eMoCA‐Thai, making it a useful tool to help improve workflow and increase accessibility.
Real-time streaming applications with multiple heterogeneous data streams have become increasingly popular especially in IoT applications; however, many issues still exist, especially in deploying ...and maintaining these large amounts of data streams. Using Spark Structured Streaming, this paper introduces a Spark Streaming framework for multiple heterogeneous data streams which allows the deployment of multiple heterogeneous data stream processing in a single Spark application; reducing deployment difficulty, coding redundancy, monitoring difficulties, and solving the problem of inefficient job queueing in multi-stream applications.
The Hadoop Distributed File System (HDFS) is an open source system which is designed to run on commodity hardware and is suitable for applications that have large data sets (terabytes). As HDFS ...architecture bases on single master (NameNode) to handle metadata management for multiple slaves (Datanode), NameNode often becomes bottleneck, especially when handling large number of small files. To maximize efficiency, NameNode stores the entire metadata of HDFS in its main memory. With too many small files, NameNode can be running out of memory. In this paper, we propose a mechanism based on Hadoop Archive (HAR), called New Hadoop Archive (NHAR), to improve the memory utilization for metadata and enhance the efficiency of accessing small files in HDFS. In addition, we also extend HAR capabilities to allow additional files to be inserted into the existing archive files. Our experiment results show that our approach can to improve the access efficiencies of small files drastically as it outperforms HAR up to 85.47%.
MongoDB is a Document based NoSQL that developed to answer the increasing demands for scalable date store in big data era. To achieve good performance, the system must be designed properly from the ...start. The performance issues like hotspot can be fatal to overall performance and various researches were focusing on applying load balancing techniques to resolve this problem. However, these techniques may not be effective for time series data, such as real-time system logs. In this paper, we propose a new data distribution algorithm based on tag aware sharding to minimize the effect of hotspot problem, especially the system with heavy writing requirements. This improves the overall performance and allow us to handle time series data effectively.
Log processing can be very challenging, especially for environments with lots of servers. In these environments, log data is large, coming at high-speed, and have various formats, the classic case of ...big data problem. This makes anomaly detection very difficult due to the fact that to get good accuracy, large amount of data must be processed in real-time. To solve this problem, this paper proposes a hybrid architecture for log anomaly detection using Apache Spark for data processing and Apache Flume for data collecting. To demonstrate the capabilities of our proposed solution, we implement a SARIMA-based anomaly detection as a case study. The experimental results clearly indicated that our proposed architecture can support log processing in large-scale environment effectively.
PDF
