DSpace collection: 期刊論文

Double Exponential Smoothing Slime Mould Algorithm For Disease Detection In Iot Healthcare System

title: Double Exponential Smoothing Slime Mould Algorithm For Disease Detection In Iot Healthcare System abstract: This paper presents an algorithm, called the double exponential smoothing slime mould algorithm (DeSSMA), which is formulated to train deep learning models for the precise detection of diseases in patients. The DeSSMA is designed by integrating the principles of double exponential smoothing with the slime mould algorithm. The parameters, including energy depletion, link lifetime (LLT), and distance, are considered by the proposed DeSSMA as objectives aimed at optimizing data routing efficiency. In the base station, a deep residual network (DRN) is trained using the proposed DeSSMA algorithm, which is utilized for disease detection following the processes of data preprocessing, augmentation, and feature selection. Finally, performance evaluation of the DeSSMA-DRN framework is conducted using metrics such as energy consumption, LLT, accuracy, sensitivity, specificity, and receiver operating characteristic. The findings reveal that the proposed framework achieved a minimal energy depletion rate of 0.412 (J), an LLT rate of 0.318, an increased accuracy rate of 0.959, a high sensitivity rate of 0.967, and a specificity rate of 0.931.

Role of Corporate Social Responsibility in the Financial Sustainability of Sports Organizations

title: Role of Corporate Social Responsibility in the Financial Sustainability of Sports Organizations abstract: Corporate social responsibility (CSR) has gradually more important for sports organizations as they seek to improve their financial sustainability. A growing number of sports organizations are recognizing the strategic value and potential benefits of CSR and are implementing CSR initiatives into their operations. The study explores the impact of CSR on the financial sustainability of sports organizations. Through a complete assessment of the literature and analysis of relevant data, the study examines the impact of various CSR initiatives, such as sponsorship and funding, charities through partnerships, organizational commitment, and stakeholder satisfaction on the financial sustainability of sports organizations. Furthermore, the study analyzes the impact of external factors, such as global economic recession and proper fund utilization on the financial sustainability of sports organizations. The findings reveal that proper implementation of CSR initiatives can positively impact the financial sustainability of sports organizations. CSR partnerships with charitable organizations can provide an effective way for sports organizations to improve their social profile, while also helping to generate revenue and financial stability. The study also highlights the significance of stakeholder satisfaction and organizational commitment towards CSR activities in improving financial sustainability. The outcome of this study underscores the importance of CSR in the financial sustainability of sports organizations and provides insights into how sports organizations can leverage CSR to improve their long-term financial sustainability while also benefiting society at large. The findings show that factors, such as sponsorship and funding, proper fund utilization, global economic recession, charities through CSR partnerships, organizational commitment to CSR activities, and stakeholder satisfaction have a significant impact on the financial improvement of sports organizations. The factor “improving social profile” doesn’t have an impact on the financial improvement of sports organizations.

Web-Based Personalized Machine Learning Recommendations to Enhance Shared Decision-Making in Prostate-Specific Antigen Screening: Randomized Controlled Trial

title: Web-Based Personalized Machine Learning Recommendations to Enhance Shared Decision-Making in Prostate-Specific Antigen Screening: Randomized Controlled Trial abstract: Background: Prostate‑specific antigen (PSA) screening involves complex trade‑offs between early detection and the risks of overdiagnosis. For older adults (aged ≥50 years), shared decision‑making (SDM) is often hindered by limited health literacy, sensory or cognitive impairments, and multimorbidity, which complicate risk comprehension. Traditional decision aids provide foundational knowledge but are often nonpersonalized. Machine learning (ML) may offer individualized recommendations, yet the psychological and behavioral effects of ML‑assisted SDM in geriatric populations remain poorly characterized. Objective: This study aimed to develop and evaluate a web‑based, ML‑driven decision aid integrated into an SDM workflow to provide personalized PSA screening recommendations and to assess its effects on decisional conflict (primary outcome), state anxiety, and decision satisfaction among middle‑aged and older men. Methods: The study followed a 2‑stage design. First, a model establishment group (n=507) was used to train and evaluate 6 ML algorithms based on clinical and values‑clarification data. A random forest model was selected for its superior performance (mean area under the curve 0.933, SD 0.350; 95% CI 0.902-0.963). Second, a randomized controlled trial was conducted with 367 participants (mean age 64.34, SD 10.30 years) randomly assigned 1:1 to the ML suggestion group (MLSG; n=185) or the control group (CG; n=182). Both groups received video‑based education, counseling, and values clarification; only the MLSG received an ML‑generated "second opinion" recommendation. Primary and secondary outcomes were assessed using the Decisional Conflict Scale (DCS), Spielberger State‑Trait Anxiety Inventory (STAI), and Satisfaction with Decision scale. Results: In the randomized controlled trial (n=367), the MLSG reported significantly lower decisional conflict than the CG (total DCS score: mean difference [MD] -3.77, 95% CI -5.55 to -1.99; Cohen d=-0.44; P<.001). The MLSG reported greater perceived support (DCS7: adjusted P=.03), more adequate advice (DCS9: adjusted P<.001), and higher decision confidence (DCS10: adjusted P=.03; DCS11: adjusted P<.001). Regarding psychological well‑being, although total anxiety scores did not differ, the MLSG reported reduced worry (STAI item 6: MD -0.98, 95% CI -1.20 to -0.76; d=-0.89; adjusted P<.001) and increased calmness (STAI item 1: MD 0.30, 95% CI 0.06-0.54; d=0.25; adjusted P=.01). Decision satisfaction was higher in the MLSG across all items (total Satisfaction with Decision score: MD -7.38, 95% CI -8.54 to -6.18; P<.001). Behavioral choices were strongly influenced by the ML recommendation: participants in the MLSG who received an "accept" recommendation were more likely to select "accept" (34/67, 50.7%) than those in the CG (44/182, 24.2%; P<.001). When the system suggested "not now," only 17.8% (21/118) chose "accept," which was lower than in the CG. Conclusions: Integrating personalized ML recommendations into SDM workflows provides emotional scaffolding for older men, reducing decisional distress and enhancing confidence without undermining autonomy. By addressing geriatric‑specific vulnerabilities through a facilitated digital interface, this ML‑driven approach complements traditional clinical consultations. These findings support the scalable integration of artificial intelligence-assisted decision support to foster patient‑centered care in aging populations.

Decoupled Detection and Category-Level 6D Pose Estimation for Robot Grasping

title: Decoupled Detection and Category-Level 6D Pose Estimation for Robot Grasping abstract: 6D object pose estimation is an essential component for robotic grasping. Most existing deep learning-based approaches focus on instance-level pose estimation, which requires prior object models and consequently limits their applicability on unseen objects in real-world scenarios. In contrast, category-level 6D pose estimation adopts Normalized Object Coordinate Space (NOCS) maps to represent intra-class object geometry, enabling pose prediction without relying on predefined object models and thus improving generalization to unseen instances. However, the original NOCS-based category-level framework typically trains NOCS prediction and object classification in a joint manner, which introduces NOCS regression error among inter-class instances with similar appearances, thereby degrading pose estimation accuracy. To address this issue, we integrate the YOLOv8 object detection with SegFormer and propose a novel Category-Level SegFormer for 6D Object Pose Estimation (CLSF-6DPE). By decoupling object classification from NOCS regression through independent learning branches, the proposed framework significantly improves pose estimation performance. Furthermore, we validate the practical feasibility of CLSF-6DPE by integrating it with a robotic gripper via the Robot Operating System (ROS) in a Real-World grasping setup. Experimental results on the CAMERA and Real-World datasets demonstrate that the proposed method achieves mAP scores of 93.8% and 81.1%, respectively. Overall, the proposed method provides a modular and effective solution for category-level pose estimation in real-world robotic grasping applications.

Integrating deep learning and groundwater dynamics for drought vulnerability assessment under climate scenarios

title: Integrating deep learning and groundwater dynamics for drought vulnerability assessment under climate scenarios abstract: Drought increasingly threatens agricultural sustainability, particularly in groundwater-dependent regions where irrigation and aquifer recharge are closely linked. Taiwan's Zhuoshui River alluvial fan exemplifies this risk: long-term intensive pumping and rising climate extremes have amplified drought vulnerability. Yet most existing drought indices treat groundwater implicitly, and many AI studies focus on groundwater prediction without translating results into integrated vulnerability metrics. This study develops an AI-driven framework to assess future drought risk from climate, groundwater, and socio-environmental drivers. Groundwater level was predicted using a hybrid Convolutional Neural Network–Backpropagation model (CNN-BP) calibrated with 22 years of basin-wide gridded precipitation, temperature, and SPI data, together with groundwater levels from 18 monitoring wells. CNN-BP outperforms a BPNN benchmark, improving the correlation coefficient by 35.85% and reducing MAE by 19.51%, enabling robust projections for 2021–2100. These groundwater forecasts are then integrated with climatic (SPI), physiographic (soil, land use, elevation, slope, distance to river) and socio-economic (population) drivers to construct the Deep Learning-based Comprehensive Drought Vulnerability Indicator (DCDVI) under SSP1-2.6 and SSP5-8.5. Scenario results indicate consistent intensification of drought vulnerability relative to the historical baseline. SSP1-2.6 yields milder drought conditions and slower groundwater decline, while SSP5-8.5 leads to stronger drying and higher vulnerability. Under SSP5-8.5, highly vulnerable areas increase from 27.31% to 41.26% by 2081–2100. Overall, DCDVI provides a scalable, climate-responsive indicator that converts AI-based groundwater forecasts into actionable vulnerability maps. The framework provides a transferable decision-support tool for drought-prone, groundwater-reliant farming systems under climate change.

A CNN-transformer framework for air quality forecasting to support aeolian dust management in river basins

title: A CNN-transformer framework for air quality forecasting to support aeolian dust management in river basins abstract: Accurately forecasting riverbed aeolian dust emissions (PM10) in complex watershed environments is a critical engineering challenge, shaped by the intricate interdependencies among hydrometeorological factors, land surface dynamics, and anthropogenic pollution sources. Traditional models often struggle to capture these nonlinear interactions, limiting their utility for real-time environmental decision-making. This study presents a novel hybrid deep learning framework—combining a 3D Convolutional Neural Network (CNN), dual 1D CNNs, and a Transformer architecture—to enhance the predictive accuracy and interpretability of PM1110 forecasts in Taiwan’s Jhuoshuei River Basin. The model harnesses the spatial feature extraction of the 3D CNN, temporal pattern recognition of the 1D CNNs, and long-range dependency modeling of the Transformer to learn complex, multiscale relationships across diverse environmental variables. Extensive quantitative and qualitative evaluations demonstrate the model’s superior performance over conventional approaches, particularly in capturing seasonal variability and the mitigating effects of water infrastructure (e.g., Jiji Weir discharge) on dust emissions. The model effectively anticipates pollution peaks, offering critical lead time for the implementation of targeted interventions such as reservoir releases or dust suppression. Beyond technical innovation, this research provides actionable insights into the dynamic coupling of atmospheric, hydrological, and operational factors. The model’s scalability and generalizability position it as a robust decision-support tool for engineers, environmental managers, and policymakers. By bridging AI-driven modeling with practical engineering applications, this study advances the field of environmental informatics and supports the development of adaptive, knowledge-based systems for sustainable air quality and watershed management.

A case study on the application of a data-driven (XGBoost) approach on the environmental and socio-economic perspectives of agricultural groundwater management

title: A case study on the application of a data-driven (XGBoost) approach on the environmental and socio-economic perspectives of agricultural groundwater management abstract: Climate-induced extreme hydrological events threaten irrigation water resources and crop production. Groundwater serves as a vital source of irrigation during periods of surface water scarcity; however, excessive and unsustainable abstraction has resulted in land subsidence. While reducing groundwater over-extraction can alleviate this issue, it may also compromise agricultural productivity, particularly during drought conditions. To address this, a reliable assessment tool is needed to balance sustainable groundwater extraction and agricultural productivity. This study develops a groundwater level prediction model using the extreme gradient boosting (XGB) algorithm, employing power consumption, precipitation, and groundwater level data as input features. Bayesian optimization was used to determine the best-fit hyperparameters, resulting in RMSE, MAE, and R² values ranging from 0.923 to 2.497 m, 0.709–2.132 m, and 0.057–0.914, respectively, during model validation. Model testing from January 2022 to June 2023 showed a strong correlation between monitored and predicted levels, indicating effective trend capture, despite slight overestimations during the dry seasons. Scenario predictions showed that a 50 % reduction in power consumption for double-crop rice led to groundwater level increases of 0.41–2.31 m in the wet season and 0.54–2.52 m in the dry season, maintaining safe thresholds. However, current fallowing subsidies recover only a fraction of the economic profit from rice production, limiting policy adoption. To improve long-term effectiveness, this study recommends institutionalizing adaptive fallowing policies, such as seasonally adjusted quotas based on real-time groundwater and rainfall indicators, and tiered subsidy schemes according to groundwater risk levels. Embedding these tools within broader agricultural governance frameworks can enhance policy responsiveness and sustainability. The proposed model supports both short-term decision-making and long-term climate-informed groundwater management by balancing environmental protection with food security and economic viability.

Advanced groundwater level forecasting with hybrid deep learning model: Tackling water challenges in Taiwan's largest alluvial fan

title: Advanced groundwater level forecasting with hybrid deep learning model: Tackling water challenges in Taiwan's largest alluvial fan abstract: Groundwater is crucial for food security and economic development, yet it faces growing threats from over-extraction and extreme weather events. The Zhuoshui River alluvial fan, Taiwan’s largest, has long served as a key water source. However, recent climate change and industrial expansion have significantly affected groundwater recharge and quality, contributing to land subsidence. Accurate forecasting of groundwater levels is essential to ensuring environmental sustainability in the region. This study presents a novel hybrid deep learning model, CNN-BP, which integrates Convolutional Neural Networks (CNN) with Backpropagation Neural Networks (BPNN) to forecast groundwater levels three days in advance at 25 monitoring stations across the Zhuoshui River alluvial fan. The CNN-BP model was benchmarked against a standalone BPNN model. Both models were trained on a dataset of 7,291 daily hydro-geo-meteorological records from 2000 to 2019, including groundwater levels, rainfall, streamflow, temperature, evaporation, and lithology. The study emphasizes comprehensive input selection, feature extraction, and hyperparameter tuning, with Random Forest utilized to filter input factors from 20 rainfall stations, thereby improving forecast accuracy and reliability. The CNN-BP model significantly outperformed the BPNN model, achieving R2 values between 0.94 and 0.98 across various stations and effectively mitigating time-delay issues. The study also explored the relationship between forecast errors and the fan’s lithological characteristics, providing valuable insights for land-use planning and groundwater management. Validation during Typhoons Haitang and Maria further demonstrated the model’s capability to predict groundwater recharge under intense rainfall conditions. By integrating environmental and social factors such as drought frequency, population density, and recharge potential, this study underscores the need for targeted water management strategies. The findings offer critical insights for future regional approaches to groundwater management, promoting sustainable practices across watersheds. Ultimately, this study serves as a valuable resource for informed decision-making in land-use planning and water resource management, advancing the sustainable utilization of groundwater in the Zhuoshui River alluvial fan.

Intelligent Urban Flood Management Using Real-Time Forecasting, Multi-Objective Optimization, and Adaptive Pump Operation

title: Intelligent Urban Flood Management Using Real-Time Forecasting, Multi-Objective Optimization, and Adaptive Pump Operation abstract: Climate-induced extreme rainfall events are increasing the intensity and frequency of flash floods, highlighting the urgent need for advanced flood management systems in climate-resilient cities. This study introduces an Intelligent Flood Control Decision Support System (IFCDSS), a novel AI-driven solution for real-time flood forecasting and automated pump operations. The IFCDSS integrates multiple advanced tools: machine learning for rapid short-term water level forecasting, NSGA-III for multi-objective optimization, the TOPSIS for robust multi-criteria decision-making, and the ANFIS for real-time pump control. Implemented in the flood-prone Zhongshan Pumping Station catchment in Taipei, the IFCDSS leveraged real-time sensor data to deliver accurate water level forecasts within five seconds for the next 10–30 min, enabling proactive and informed operational responses. Performance evaluations confirm the system’s scientific soundness and practical utility. Specifically, the ANFIS achieved strong accuracy (R2 = 0.81), with most of the prediction errors being limited to a single pump unit. While the conventional manual operations slightly outperformed the IFCDSS in minimizing flood peaks—due to their singular focus—the IFCDSS excelled in balancing multiple objectives: flood mitigation, energy efficiency, and operational reliability. By simultaneously addressing these dimensions, the IFCDSS provides a robust and adaptable framework for urban environments. This study highlights the transformative potential of intelligent flood control to enhance urban resilience and promote sustainable, climate-adaptive development.

AI-driven weather downscaling for smart agriculture using autoencoders and transformers

title: AI-driven weather downscaling for smart agriculture using autoencoders and transformers abstract: Artificial Intelligence (AI) is reshaping agriculture by driving smarter, data-driven practices that enhance regional weather forecasting and support proactive, informed decision-making. Advances in Big Data, IoT, Remote Sensing, and Machine Learning are accelerating this transformation, with Transformer architectures increasingly pivotal in refining agricultural management strategies, especially in Taiwan. In this study, we develop a hybrid Convolutional Autoencoder and LSTM-based Transformer Network (CAE-LSTMT) to downscale six-hour simulation data into precise hourly forecasts, validated using 55,538 temperature and relative humidity records (2020–2023) from Taiwan’s Jhuoshuei River basin, provided by the Central Weather Administration (CWA). The model was trained (70 %), validated (10 %), and tested (20 %) to optimize its configuration and performance. This CAE-LSTMT model substantially enhances spatiotemporal weather forecast resolution, transforming six-hour regional data into hourly forecasts with improved accuracy. It yields temperature forecast gains of 5.66 % to 20.39 % and relative humidity improvements of 8.05 % to 12.76 %, with reduced forecast biases compared to traditional LSTM models. The model demonstrates exceptional accuracy in vapor pressure deficit (VPD) predictions, achieving mean absolute errors (MAE) between 0.15 to 0.21 kPa across regions and 0.16 to 0.20 kPa seasonally, significantly outperforming the CWA model. Accurate VPD forecasts allow farmers to manage irrigation and minimize crop stress, directly supporting plant health and yield optimization. For heat index classification, the model achieves up to 96 % ACCURACY, with mean absolute percentage errors (MAPE) of 4 % to 23 %, significantly exceeding the CWA model’s ACCURACY range of 35 % to 79 % and MAPE of 29 % to 70 %. This high precision in heat index forecasting empowers farmers to protect crops and livestock against heat stress. By extracting critical features from high-dimensional data, the CAE-LSTMT model advances environmental downscaling for multi-site, multi-horizon weather data, showing significant promise for Smart Agriculture and Health Advisory Systems. This approach offers precise, actionable forecasts, optimizing agricultural practices and reducing climate-related risks, underscoring its impact on sustainable agricultural and environmental management.

Flood resilience through hybrid deep learning: Advanced forecasting for Taipei's urban drainage system

title: Flood resilience through hybrid deep learning: Advanced forecasting for Taipei's urban drainage system abstract: The escalating impacts of climate change have intensified extreme rainfall events, placing urban drainage systems under unprecedented pressure and increasing flood risks. Addressing these challenges requires advanced flood mitigation strategies, optimized sewer operations, and responsive disaster management. This study leverages knowledge graphs to integrate diverse data sources, providing a comprehensive perspective on flood dynamics, and applies deep learning models within a Real-Time Urban Drainage Early Warning System to enhance flood management at Taipei City's Zhongshan Pumping Station in Taiwan. We proposed deep learning models, specifically Convolutional Neural Networks combined with Back Propagation Neural Networks (CNN-BP), to make multi-input multi-output multi-step (MIMOMS) forecasts on sewer water levels at intervals from 10 to 40 min (T+1 to T+4) and MIMO forecasts on the pumping station's internal (forebay) and external (river) water levels at intervals from 10 to 60 min (T+1 to T+6). The CNN-BP model exhibited superior forecast accuracy, reaching an R2 (RMSE) of 0.97 (0.08m) at T+1 for sewer water levels and an R2 (RMSE) of 0.99 (0.06m) at T+1 for both internal and external water levels. These results highlight CNN-BP's capability to accurately capture water level trends, ensuring reliable real-time responsiveness, especially during intense and sudden rainfall events. The CNN-BP's high predictive accuracy enables enhanced pump operations, strengthens early warning systems, and fosters intelligent flood control practices crucial for effective environmental management.

Attention Distribution-Aware Softmax for NPU-Accelerated On-Device Inference of LLMs: An Edge-Oriented Approximation Design

title: Attention Distribution-Aware Softmax for NPU-Accelerated On-Device Inference of LLMs: An Edge-Oriented Approximation Design abstract: Low-power NPUs enable on-device LLM inference through efficient integer and fixed-point algebra, yet their lack of native exponential support makes Transformer softmax a critical performance bottleneck. Existing NPU kernels approximate e^x using uniform piecewise polynomials to enable O(1) SIMD indexing, but this wastes computation by applying high-degree arithmetic indiscriminately in every segment. Conversely, fully adaptive approaches maximize statistical fidelity but introduce pipeline stalls due to comparator-based boundary search. To bridge this gap, we propose an attention distribution-aware softmax that uses Particle Swarm Optimization (PSO) to define non-uniform segments and variable polynomial degrees, prioritizing finer granularity and lower arithmetic complexity in attention-dense regions. To ensure efficiency, we snap boundaries into a 128-bin LUT, enabling O(1) retrieval of segment parameters without branching. Inference measurements show that this favors low-degree execution, minimizing exp-kernel overhead. Using TinyLlama-1.1B-Chat as a testbed, the proposed weighted design reduces cycles per call exp kernel (CPC) by 18.5% versus an equidistant uniform Degree-4 baseline and 13.1% versus uniform Degree-3, while preserving ranking fidelity. These results show that grid-snapped, variable-degree approximation can improve softmax efficiency while largely preserving attention ranking fidelity, enabling accurate edge LLM inference.

感謝表現として多文化多言語社会の台湾にトランスフォーメーションした村上春樹の「小確幸」ランブクピティヤ　ディヌーシャ編

title: 感謝表現として多文化多言語社会の台湾にトランスフォーメーションした村上春樹の「小確幸」ランブクピティヤ　ディヌーシャ編

洋行者漱石の文明開化との苦闘と回帰昇華

title: 洋行者漱石の文明開化との苦闘と回帰昇華

パートナーシップの観点から見た『ノルウェイの森』の男同士の関係変容―「死は生の対極としてではなく、その一部として存在している」の言葉に注目して―

title: パートナーシップの観点から見た『ノルウェイの森』の男同士の関係変容―「死は生の対極としてではなく、その一部として存在している」の言葉に注目して―

「死の文学」と言われた村上春樹の1980年代までの創作群に築いた死生観―『ダンス・ダンス・ダンス』とその前の短篇集を中心に―

title: 「死の文学」と言われた村上春樹の1980年代までの創作群に築いた死生観―『ダンス・ダンス・ダンス』とその前の短篇集を中心に―

日語翻譯教育之於AI，AI之於日語翻譯教育的課題：AI賦能學習者創新自主學習的活教材

title: 日語翻譯教育之於AI，AI之於日語翻譯教育的課題：AI賦能學習者創新自主學習的活教材 abstract: 「後AI時代」一詞出現，代表AI被廣泛通用現象業已形成。又大腦學習為主的教師與LLM學習為主的數位原生世代學習者間出現的學習模式的世代差異，教師如何克服兩者間差異性，而進行教與學，乃是本稿撰寫的動機。當著眼於日語學習5項技能中的「譯」，理念貫穿以下4步驟，先以教師傳承豐富的教學經驗的點出長年以來學生容易犯錯之處。再由教師介紹學習者有利改善的AI工具，且與學生並肩作戰找出對策，解決學生學習的問題。繼之借力AI讓學生於課外時間進行自主學習。最後由教師認證自主學習成效，給予評價。依此理念，借力AI為應用策略，師生聯袂將自製日語翻譯自主學習教材，廣為推行。如此地使得AI賦能學習者自製自主學習的日語翻譯活教材，足以讓日語翻譯教育之於AI，AI之於日語翻譯教育的課題，得以永續發展。

A CNN-transformer framework for air quality forecasting to support aeolian dust management in river basins

High-spatiotemporal-resolution PM2.5 forecasting by hybrid deep learning models with ensembled massive heterogeneous monitoring data

title: High-spatiotemporal-resolution PM2.5 forecasting by hybrid deep learning models with ensembled massive heterogeneous monitoring data abstract: High-resolution real-time air quality forecasting can alert decision-makers and residents about forthcoming air pollution events and refine air quality management. The Environmental Protection Administration in Taiwan has deployed numerous low-cost air quality microsensors near industrial zones lately to facilitate local air quality monitoring. Nevertheless, the frequent occurrence of missing sensor data due to problems of mobile transmission, frontend/backend device malfunction, or other unforeseen issues would raise difficulty in making quick responses to air pollution incidents. This study proposed a hybrid deep learning model (AE-CNN-BP) collaborating an Autoencoder (AE), a Convolutional Neural Network (CNN), and a Back Propagation Neural Network (BPNN) to effectively extract crucial features from big data for making successive high-spatiotemporal-resolution forecasts of PM2.5 concentrations 4 h ahead. The proposed model was trained and tested in three industrial zones densely installed with microsensors in Kaohsiung City of Taiwan. A high pollution incident was selected to evaluate model performance. The results show that the proposed model could reliably produce nice high-spatiotemporal-resolution forecasts for 12 air quality monitoring stations and 485 microsensors, with Coefficient of Determination (R2) values and Root Mean Squared Error (RMSE) of 0.82 (0.76) and 11.05 (12.75) μg/m3 in the training (testing) stage, respectively. For the selected incident, the Mean Absolute Percentage Error (MAPE) values of the proposed model were 22.3% and 27.1% at T+1 and T+4, respectively. This study demonstrates that the proposed deep learning model based on ensemble datasets of sparsely distributed monitoring stations and densely deployed microsensors can offer reliable high-spatiotemporal-resolution air quality forecasts, benefiting environmental studies and informed policymaking by accounting for local-scale variations in PM2.5 concentrations.

Develop a hybrid machine learning model for promoting microbe biomass production

title: Develop a hybrid machine learning model for promoting microbe biomass production abstract: Since the cultivation condition of microbe biomass production (mycelia yield) involves a variety of factors, it’s a laborious process to obtain the optimal cultivation condition of Antrodia cinnamomea (A. cinnamomea). This study proposed a hybrid machine learning approach (i.e., ANFIS-NM) to identify the potent factors and optimize the cultivation conditions of A. cinnamomea based on a 32 fractional factorial design with seven factors. The results indicate that the ANFIS-NM approach successfully identified three key factors (i.e., glucose, potato dextrose broth, and agar) and significantly boosted mycelia yield. The interpretability of ANFIS rules made the cultivation conditions visually interpretable. Subsequently, a three-factor five-level central composite design was used to probe the optimal yield. This study demonstrates the proposed hybrid machine learning approach could significantly reduce the time consumption in laboratory cultivation and increase mycelia yield that meets SDGs 7 and 12, hitting a new milestone for biomass production.

Advancing climate-resilient flood mitigation: Utilizing transformer-LSTM for water level forecasting at pumping stations

title: Advancing climate-resilient flood mitigation: Utilizing transformer-LSTM for water level forecasting at pumping stations abstract: Proactive management of pumping stations using artificial intelligence (AI) technology is vital for effectively mitigating the impacts of flood events caused by climate change. Accurate water level forecasts are pivotal in advancing the intelligent operation of pumping stations. This study proposed a novel Transformer-LSTM model to offer accurate multi-step-ahead forecasts of the flood storage pond (FSP) and river water levels for the Zhongshan pumping station in Taipei, Taiwan. A total of 19,647 ten-minute-based datasets of pumping operation and storm sewer, FSP, and river water levels were collected between 2014 and 2020 and further divided into training (70 %), validation (10 %), and test (20 %) datasets for model construction. The results demonstrate that the proposed model dramatically outperforms benchmark models by producing more accurate and reliable water level forecasts at 10-minute (T + 1) to 60-minute (T + 6) horizons. The proposed model effectively enhances the connections between input factors through the Transformer module and increases the connectivity across consecutive time series using the LSTM module. This study reveals interconnected dynamics among pumping operation and storm sewer, FSP, and river water levels, enhancing flood management. Understanding these dynamics is crucial for effective execution of management strategies and infrastructure revitalization against climate impacts. The Transformer-LSTM model's forecasts encourage water practices, resilience, and disaster risk reduction for extreme weather events.

An Artificial Intelligence-Powered Environmental Control System for Resilient and Efficient Greenhouse Farming

title: An Artificial Intelligence-Powered Environmental Control System for Resilient and Efficient Greenhouse Farming abstract: The rise in extreme weather events due to climate change challenges the balance of supply and demand for high-quality agricultural products. In Taiwan, greenhouse cultivation, a key agricultural method, faces increasing summer temperatures and higher operational costs. This study presents the innovative AI-powered greenhouse environmental control system (AI-GECS), which integrates customized gridded weather forecasts, microclimate forecasts, crop physiological indicators, and automated greenhouse operations. This system utilizes a Multi-Model Super Ensemble (MMSE) forecasting framework to generate accurate hourly gridded weather forecasts. Building upon these forecasts, combined with real-time in-greenhouse meteorological data, the AI-GECS employs a hybrid deep learning model, CLSTM-CNN-BP, to project the greenhouse’s microclimate on an hourly basis. This predictive capability allows for the assessment of crop physiological indicators within the anticipated microclimate, thereby enabling preemptive adjustments to cooling systems to mitigate adverse conditions. All processes run on a cloud-based platform, automating operations for enhanced environmental control. The AI-GECS was tested in an experimental greenhouse at the Taiwan Agricultural Research Institute, showing strong alignment with greenhouse management needs. This system offers a resource-efficient, labor-saving solution, fusing microclimate forecasts with crop models to support sustainable agriculture. This study represents critical advancements in greenhouse automation, addressing the agricultural challenges of climate variability.

深度學習：環境資料數位化的應用

title: 深度學習：環境資料數位化的應用 abstract: 深度學習能精準捕捉資料中的非線性特徵和劇烈變化，實現高精度預測。本研究將其應用於多個環境議題，包括全臺灣的空氣污染、臺北市的下水道水位預測及溫室微氣候預測。空污預測使用了全臺灣環保署測站的歷史資料，涵蓋六個污染因子和兩個氣象因子，模型引入了注意力機制，成功解決傳統深度學習的梯度消失問題，顯著提升未來72小時的預測精度，誤差（root-mean-square error, RMSE）在8.5至13.2 μg/m^3之間。在臺北市的下水道水位預測中，本研究採用了DNN-AE模型，能穩定預測未來10至60分鐘水位。在所有DNN-AE模型中，C-AE模型在結構上更具優勢，卷積層能有效提取時間特徵，特別是在處理時間變化資料時，能捕捉更細緻的趨勢，顯著提高預測精度，誤差（RMSE）在T+1至T+6為0.21至0.51 m，預測結果最佳且穩定。對於溫室微氣候預測，本研究使用XGBoost對彰化伸港各微氣候因子進行特徵篩選，分析了溫度、相對濕度和光照強度等關鍵影響因素。結果顯示影響溫度、相對濕度和光照強度的主要特徵分別為溫度、短波輻射（日射量）和RH，反映了捲簾和遮蔽系統對溫室內部微氣候的影響。結果顯示，ANFIS模型在溫度預測方面表現最佳，R^2值超過0.8，CNN則在相對濕度與光照強度的預測上表現良好。綜合這些應用，臺灣的環境預測技術將能進一步推動智慧城市的發展，邁向更高階的數位化未來。

Watershed groundwater level multistep ahead forecasts by fusing convolutional-based autoencoder and LSTM models

title: Watershed groundwater level multistep ahead forecasts by fusing convolutional-based autoencoder and LSTM models abstract: The development of deep learning-based groundwater level forecast models can tackle the challenge of high dimensional groundwater dynamics, predict groundwater variation trends accurately, and manage groundwater resources effectively, thereby contributing to sustainable water resources management. This study proposed a novel ConvAE-LSTM model, which fused a Convolutional-based Autoencoder model (ConvAE) and a Long Short-Term Memory Neural Network model (LSTM), to provide accurate spatiotemporal groundwater level forecasts over the next three months. The HBV-light and LSTM models are chosen as benchmarks. An ensemble of point data and the corresponding derived images concerning the past (observations) and the future (forecasts from a conceptual model) of groundwater levels at 33 groundwater wells in Jhuoshuei River basin of Taiwan between 2000 and 2019 constituted the case study. The findings showcase the effectiveness of the ConvAE-LSTM model in extracting crucial features from both point and imagery datasets. This model successfully establishes spatiotemporal dependencies between regional images and groundwater level data over diverse time frames, leading to accurate multi-step-ahead forecasts of groundwater levels. Notably, the ConvAE-LSTM model exhibits a substantial improvement, with the R-squared values showing an increase of more than 18%, 22%, and 49% for the R1, R2, and R3 regions, respectively, compared to the HBV-light model. Additionally, it outperforms the LSTM model in this regard. This study represents a noteworthy milestone in environmental modeling, offering key insights for designing sustainable groundwater management strategies to ensure the long-term availability of this vital resource.

Advanced groundwater level forecasting with hybrid deep learning model: Tackling water challenges in Taiwan’s largest alluvial fan

title: Advanced groundwater level forecasting with hybrid deep learning model: Tackling water challenges in Taiwan’s largest alluvial fan abstract: Groundwater is crucial for food security and economic development, yet it faces growing threats from over-extraction and extreme weather events. The Zhuoshui River alluvial fan, Taiwan’s largest, has long served as a key water source. However, recent climate change and industrial expansion have significantly affected groundwater recharge and quality, contributing to land subsidence. Accurate forecasting of groundwater levels is essential to ensuring environmental sustainability in the region. This study presents a novel hybrid deep learning model, CNN-BP, which integrates Convolutional Neural Networks (CNN) with Backpropagation Neural Networks (BPNN) to forecast groundwater levels three days in advance at 25 monitoring stations across the Zhuoshui River alluvial fan. The CNN-BP model was benchmarked against a standalone BPNN model. Both models were trained on a dataset of 7,291 daily hydro-geo-meteorological records from 2000 to 2019, including groundwater levels, rainfall, streamflow, temperature, evaporation, and lithology. The study emphasizes comprehensive input selection, feature extraction, and hyperparameter tuning, with Random Forest utilized to filter input factors from 20 rainfall stations, thereby improving forecast accuracy and reliability. The CNN-BP model significantly outperformed the BPNN model, achieving R2 values between 0.94 and 0.98 across various stations and effectively mitigating time-delay issues. The study also explored the relationship between forecast errors and the fan’s lithological characteristics, providing valuable insights for land-use planning and groundwater management. Validation during Typhoons Haitang and Maria further demonstrated the model’s capability to predict groundwater recharge under intense rainfall conditions. By integrating environmental and social factors such as drought frequency, population density, and recharge potential, this study underscores the need for targeted water management strategies. The findings offer critical insights for future regional approaches to groundwater management, promoting sustainable practices across watersheds. Ultimately, this study serves as a valuable resource for informed decision-making in land-use planning and water resource management, advancing the sustainable utilization of groundwater in the Zhuoshui River alluvial fan.

A speech enhancement method using Fast Fourier Transform and Convolutional Autoencoder

title: A speech enhancement method using Fast Fourier Transform and Convolutional Autoencoder abstract: This paper addresses the reconstruction of audio signals from degraded measurements. We propose a lightweight model that combines the discrete Fourier transform with a Convolutional Autoencoder (FFT-ConvAE), which enabled our team to achieve second place in the Helsinki Speech Challenge 2024. Our results, together with those of other teams, demonstrate the potential of simple methods for effective speech reconstruction.

Intelligent Urban Flood Management Using Real-Time Forecasting, Multi-Objective Optimization, and Adaptive Pump Operation

Interactive urban building energy modelling with functional mockup interface of a local residential building stock

title: Interactive urban building energy modelling with functional mockup interface of a local residential building stock abstract: The transformation towards a low carbon energy system requires municipalities to improve their local building stock. Urban building energy modelling (UBEM) is an emerging tool to support municipalities in shaping the necessary strategies by estimating energy demand with high spatial and temporal resolution. This study proposes a Functional Mockup Interface (FMI)-based UBEM that enables interactive capacities to simulate diverse environmental conditions without reinitialisation. The FMI-based approach allows to couple the building energy simulation EnergyPlus with external models. These capacities were tested on a real-world example in the German city of Wuppertal with urban microclimate data. The results are estimates of sub-hourly energy demand based on the adjacent environmental conditions of each building. In order to ameliorate the applicability, the FMI-based UBEM is further enriched by incorporating an automatic procedure to derive 3D building models, displaying high geometrical fidelity, from city-wide point clouds through the screened Poisson surface reconstruction algorithm. The study area contains 5736 residential buildings. A diverse residential building stock was modelled on the basis of the EU project TABULA. To demonstrate the functionality of the proposed UBEM, a demand response scenario was constructed with microclimate data and heat pumps instead of other heating and hot water systems. The capacity of UBE-FMI to dynamically change parameters (e.g. thermostat setpoint) in the building models can benefit the evaluation of demand response strategies and its potential to shed peak loads. In comparison with reference studies, UBE-FMI produced reasonable estimates of energy demand. The 3D building models and simulation results using “live” weather are visualized by a web-interface, which is implemented with the geospatial 3D framework NASA WorldWind.

Seamless integration of convolutional and back-propagation neural networks for regional multi-step-ahead PM2.5 forecasting

title: Seamless integration of convolutional and back-propagation neural networks for regional multi-step-ahead PM2.5 forecasting abstract: The fine particulate matter (e.g. PM2.5) gains an increasing concern of human health deterioration. Modelling PM2.5 concentrations remains a substantial challenge due to the limited understanding of the dynamic processes as well as uncertainties residing in the emission data and their projections. This study proposed a hybrid model (CNN-BP) engaging a Convolutional Neural Network (CNN) and a Back Propagation Neural Network (BPNN) to make accurate PM2.5 forecasts for multiple stations at multiple horizons at the same time. The hourly datasets of six air quality and two meteorological factors collected from 73 air quality monitoring stations in Taiwan during 2017 formed the case study. A total of 639,480 hourly datasets were collected and allocated into training (409,238, 64%), validation (102,346, 16%), and testing (127,896, 20%) stages. The forecasts of PM2.5 concentrations were first characterized as a function of air quality and meteorological variables. Then the proposed CNN-BP approach effectively learned the dominant features of input data and simultaneously produced accurate regional multi-step-ahead PM2.5 forecasts (73 stations; t+1−t+10). The results demonstrate that the proposed CNN-BP model is remarkably superior to the BPNN, the random forest and the long short term memory neural network models owing to its higher forecast accuracy and excellence in creating reliable regional multi-step-ahead PM2.5 forecasts. Besides, the CNN-BP model not only has the power to cope with the curse of dimensionality by adequately handling heterogeneous inputs with relatively large time-lags but also has the capability to explore different PM2.5 mechanisms (local emission and transboundary transmission) for the five regions (R1-R5) and the whole Taiwan. This study shows that multi-site (regional) and multi-horizon forecasting can be achieved by exactly one model (i.e. the proposed CNN-BP model), hitting a new milestone. Therefore, the CNN-BP model can facilitate real-time PM2.5 forecast service and the forecasts can be made publicly available online.

Integrate deep learning and physically-based models for multi-step-ahead microclimate forecasting

title: Integrate deep learning and physically-based models for multi-step-ahead microclimate forecasting abstract: Precision agriculture control systems count on reliable and accurate microclimate forecasts to maintain environmental suitability for crop growth. However, IoT devices adopted to monitor microclimate are expensive to people in developing countries. This study proposed a hybrid deep learning model (ConvLSTM*CNN-BP) without using IoT data to produce accurate multi-horizon and multi-factor (greenhouse internal temperature, relative humidity, and photosynthetically active radiation) forecasts simultaneously. The proposed model fused a convolutional-based long short term memory neural network (ConvLSTM), a convolutional neural network (CNN), and a backpropagation neural network (BPNN). Model construction involved an ensemble of gridded 6-hour-ahead meteorological forecasts from the STMAS-WRF model and 3-hour-ahead greenhouse internal temperature simulated by a physically-based model at a 10-min scale. Another deep learning model (CNN*LSTM*Stacked LSTM-BP) using IoT data was established for comparison purpose. The experimental results on two greenhouses located in Central Taiwan indicated that the proposed model (non-IoT) and the benchmark model (with IoT) produced similar forecast performances on greenhouse internal temperature, relative humidity, and photosynthetically active radiation. Moreover, the proposed model with illustrious abilities of noise removal and feature extraction could provide satisfactory forecast accuracy. The proposed deep learning approach hits a milestone in multi-horizon and multi-factor forecasting on microclimate, which significantly supports farmers, especially in developing countries, in reducing the installation and maintenance costs of IoT devices for monitoring purpose.

Deep neural networks for spatiotemporal PM2.5 forecasts based on atmospheric chemical transport model output and monitoring data

title: Deep neural networks for spatiotemporal PM2.5 forecasts based on atmospheric chemical transport model output and monitoring data abstract: Reliable long-horizon PM2.5 forecasts are crucial and beneficial for health protection through early warning against air pollution. However, the dynamic nature of air quality makes PM2.5 forecasts at long horizons very challenging. This study proposed a novel machine learning-based model (MCNN-BP) that fused multiple convolutional neural networks (MCNN) with a back-propagation neural network (BPNN) for making spatiotemporal PM2.5 forecasts for the next 72 h at 74 stations covering the whole Taiwan simultaneously. Model configuration involved an ensemble of massive hourly air quality and meteorological monitoring datasets and the existing publicly-available PM2.5 simulated (forecasted) datasets from an atmospheric chemical transport (ACT) model. The proposed methodology collaboratively constructed two CNNs to mine the observed data (the past) and the forecasted data from ACT (the future) separately. The results showed that the MCNN-BP model could significantly improve the accuracy of spatiotemporal PM2.5 forecasts and substantially reduce the forecast biases of the ACT model. We demonstrated that the proposed MCNN-BP model with effective feature extraction and good denoising ability could overcome the curse of dimensionality and offer satisfactory regional long-horizon PM2.5 forecasts. Moreover, the MCNN-BP model has considerably shorter computational time (5 min) and lower computational load than the compute-intensive ACT model. The proposed approach hits a milestone in multi-site and multi-horizon forecasting, which significantly contributes to early warning against regional air pollution.

Real-time image-based air quality estimation by deep learning neural networks

title: Real-time image-based air quality estimation by deep learning neural networks abstract: Air quality profoundly impacts public health and environmental equity. Efficient and inexpensive air quality monitoring instruments could be greatly beneficial for human health and air pollution control. This study proposes an image-based deep learning model (CNN−RC) that integrates a convolutional neural network (CNN) and a regression classifier (RC) to estimate air quality at areas of interest through feature extraction from photos and feature classification into air quality levels. The models were trained and tested on datasets with different combinations of the current image, the baseline image, and HSV (hue, saturation, value) statistics for increasing model reliability and estimation accuracy. A total of 3549 hourly air quality datasets (including photos, PM2.5, PM10, and the air quality index (AQI)) collected at the Linyuan air quality monitoring station of Kaohsiung City in Taiwan constituted the case study. The main breakthrough of this study is to timely produce an accurate image-based estimation of several pollutants simultaneously by using only one single deep learning model. The test results show that estimation accuracy in terms of R2 for PM2.5, PM10, and AQI based on daytime (nighttime) images reaches 76% (83%), 84% (84%), and 76% (74%), respectively, which demonstrates the great capability of our method. The proposed model offers a promising solution for rapid and reliable multi-pollutant estimation and classification based solely on captured images. This readily scalable measurement approach could address major gaps between air quality data acquired from expensive instruments worldwide.

AI-driven weather downscaling for smart agriculture using autoencoders and transformers

ResNet-SE-CBAM Siamese Networks for Few-Shot and Imbalanced PCB Defect Classification

title: ResNet-SE-CBAM Siamese Networks for Few-Shot and Imbalanced PCB Defect Classification abstract: Defect detection in mass production lines often involves small and imbalanced datasets, necessitating the use of few-shot learning methods. Traditional deep learning-based approaches typically rely on large datasets, limiting their applicability in real-world scenarios. This study explores few-shot learning models for detecting product defects using limited data, enhancing model generalization and stability. Unlike previous deep learning models that require extensive datasets, our approach effectively performs defect detection with minimal data. We propose a Siamese network that integrates Residual blocks, Squeeze and Excitation blocks, and Convolution Block Attention Modules (ResNet-SE-CBAM Siamese network) for feature extraction, optimized through triplet loss for embedding learning. The ResNet-SE-CBAM Siamese network incorporates two primary features: attention mechanisms and metric learning. The recently developed attention mechanisms enhance the convolutional neural network operations and significantly improve feature extraction performance. Meanwhile, metric learning allows for the addition or removal of feature classes without the need to retrain the model, improving its applicability in industrial production lines with limited defect samples. To further improve training efficiency with imbalanced datasets, we introduce a sample selection method based on the Structural Similarity Index Measure (SSIM). Additionally, a high defect rate training strategy is utilized to reduce the False Negative Rate (FNR) and ensure no missed defect detections. At the classification stage, a K-Nearest Neighbor (KNN) classifier is employed to mitigate overfitting risks and enhance stability in few-shot conditions. The experimental results demonstrate that with a good-to-defect ratio of 20:40, the proposed system achieves a classification accuracy of 94% and an FNR of 2%. Furthermore, when the number of defective samples increases to 80, the system achieves zero false negatives (FNR = 0%). The proposed metric learning approach outperforms traditional deep learning models, such as parametric-based YOLO series models in defect detection, achieving higher accuracy and lower miss rates, highlighting its potential for high-reliability industrial deployment.

Efficient License Plate Alignment and Recognition Using FPGA‑Based Edge Computing

title: Efficient License Plate Alignment and Recognition Using FPGA‑Based Edge Computing abstract: Efficient and accurate license plate recognition (LPR) in unconstrained environments remains a critical challenge, particularly when confronted with skewed imaging angles and the limited computational capabilities of edge devices. In this study, we propose a high-performance, FPGA-based license plate alignment and recognition (LPAR) system to address these issues. Our LPAR system integrates lightweight deep learning models, including YOLOv4-tiny for license plate detection, a refined convolutional pose machine (CPM) for pose estimation and alignment, and a modified LPRNet for character recognition. By restructuring the pose estimation and alignment architectures to enhance the geometric correction of license plates and adding channel and spatial attention mechanisms to LPRNet for better character recognition, the proposed LPAR system improves recognition accuracy from 88.33% to 95.00%. The complete pipeline achieved a processing speed of 2.00 frames per second (FPS) on a resource-constrained FPGA platform, demonstrating its practical viability for real-time deployment in edge computing scenarios.

Multi-objective mathematical model for optimal wind turbine placement in wind farm under uncertainty

title: Multi-objective mathematical model for optimal wind turbine placement in wind farm under uncertainty abstract: The main objective of this research is to introduce three energy risk management models grounded in optimization techniques for the strategic placement of wind turbines, considering wake effects and uncertainties in wind speed and direction. For this purpose, wind speed and direction data are gathered, and Monte Carlo simulation is employed to model the uncertainties. Subsequently, the risk management models undergo optimization using Non-Dominated Sorting Genetic Algorithm II (NSGA-II), Pareto envelope-based selection algorithm II (PESA-II), and Multi-Objective Particle Swarm Optimization (MOPSO) algorithms. Findings reveal that the wind farm’s maximum power output reaches approximately 5.8 megawatts across all three algorithms and optimal turbine placements. A risk assessment was conducted using a tenth percentile criterion, revealing a significant production risk within the study area, with production falling below 1.8 megawatts in 90 % of cases. Regarding the performance evaluation of the algorithms across all three models, superior performance in terms of solution proximity to the ideal solution is exhibited by PESA-II, while enhanced diversity and solution spread compared to the other algorithms are demonstrated by NSGA-II.

Hybrid Hierarchical Attention Network-Hierarchical Deep Learning for Text Classification in Opinion Mining

title: Hybrid Hierarchical Attention Network-Hierarchical Deep Learning for Text Classification in Opinion Mining abstract: In general, opinion mining indicates the process of evaluating the opinions of people on several topics that are accessible in text form. It is an important aspect of natural language processing as it sets up the effective planning and decision-making for businesses and users. Opinion mining can be performed more effectively and conveniently by initially carrying out subjectivity recognition, which entails recognizing the text as objective or subjective. This research comprises various steps, like preprocessing, feature extraction, data augmentation and opinion mining. The complete procedure was implemented in the Spark framework that utilizes a master–slave framework. The preprocessing step is done with methods, such as stop-word removal, stemming, and lemmatization. Afterwards, feature extraction is done by extracting sentiWordNet features and statistical features that involve capitalized words, exclamation marks, and hashtags. Followed by the data augmentation, the opinion mining phase uses a HAN–HDLTex approach proposed by the combination of HAN and HDLTex architectures. The experimentation is done for the proposed HAN–HDLTex model that shows better accuracy with a rate of 0.949, sensitivity with a rate of 0.969, and specificity with a rate of 0.939.

Skill Prediction and Player Re-Identification Using Serial Exponential Ring Toss Game-Based Optimization Based Deep Maxout Network

title: Skill Prediction and Player Re-Identification Using Serial Exponential Ring Toss Game-Based Optimization Based Deep Maxout Network abstract: Over the last decade, the eSports industry has experienced significant growth. World-class eSports players now enter contracts with a team, follow a strict training regimen, and compete in tournaments. Just like conventional athletes, most eSports competitors suffer injuries that deeply affect their performance or would prevent them from training or competing. Moreover, the accuracy of performance predictions in existing studies is often constrained by insufficient prediction models. Thus, in this work, an advanced optimization-enabled deep learning approach is modeled for player re-identification and skill prediction. Primarily, the data collected from the eSports Sensor Dataset undergoes a feature selection process. This feature selection is done by using Serial Exponential-Ring TossGame-Based Optimization (SeEXP-RTGBO) method, which combines Serial Exponential Moving Average concept with the Ring Toss Game-Based Optimization (RTGBO) technique. Then, feature fusion is executed by using Morisita's overlap index and Bi-LSTM attention. The main aim of feature fusion is to enhance model performance by combining complementary information from multiple features, which leads to more accurate and robust predictions. Subsequently, skill predictions and player re-identification are established by using Deep Maxout Network (DMN), which is optimized using the proposed SeEXP-RTGBO. Ultimately, an empirical assessment is conducted by evaluating various metrics, including accuracy, sensitivity, and specificity. In this context, the presented approach achieved an accuracy of 0.959, sensitivity of 0.961, and specificity of 0.956, thereby demonstrating superiority over previously established models.

Deploying a Skeleton-Based Video Anomaly Detection System on Edge Devices for Human Activity Surveillance

title: Deploying a Skeleton-Based Video Anomaly Detection System on Edge Devices for Human Activity Surveillance abstract: Recent advances in embedded computing have enabled edge devices to run AI models more efficiently, sparking interest in deploying video anomaly detection (VAD) systems for smart surveillance. However, practical implementation requires a careful balance between detection accuracy and computational efficiency. This letter proposes a novel and lightweight anomaly scoring model that integrates a normalizing flow with a multi-scale spatial temporal graph convolutional network (stGCN). The proposed model supports both unsupervised and supervised modes. To evaluate its deployment feasibility, we implement the full VAD pipeline—including YOLOv8n-Pose, BoT-SORT, and the proposed scoring model—on a Raspberry Pi 5. Experimental results demonstrate that our method achieves AUC scores of 86.2% and 72.2% on the ShanghaiTech and UBnormal datasets for unsupervised VAD, respectively, and an AUC score of 82.4% for supervised VAD on the UBnormal dataset, outperforming state-of-the-art methods.

A novel approach for Supply Chain Shipment Pricing Prediction using Temporal Convolutional Network- Residual Neural Network

title: A novel approach for Supply Chain Shipment Pricing Prediction using Temporal Convolutional Network- Residual Neural Network abstract: The supply chain comprises an interconnected system of warehouses, suppliers, shipping companies, distribution hubs, carriers, and logistics firms collaborating to facilitate the progression and commercialization of a product until its final handover to the ultimate consumer. Moreover, efficiently managing overseas supply chains necessitates precise forecasting of shipping times, as it is a serious aspect of operations and advanced information systems. Nonetheless, the feasibility of generating real-time Global Positioning System data and employing optimization methods for short-term and long-term shipping prediction remains an important challenge. Thus, this study develops a novel approach for the supply chain shipment pricing prediction using a hybrid deep learning approach. At first, pre-processing is executed by data normalization and data transformation. Subsequently, feature fusion is performed by Atkinson index and Double Exponential Dung beetle Optimizer (DEDBO) algorithm, that is a combination of Double Exponential Smoothing (DES) and Dung beetle Optimizer (DBO). Ultimately, supply chain shipment prediction is executed by employing the Temporal Convolutional Network- Residual Neural Network (TCN-RNN), which is a combination of TCN and RNN models. The experimentation evaluation shows that DEDBO-based TCN-RNN attains minimal MSE, RMSE, MAE and MAPE with values of 0.0001, 0.0104, 0.0054 and 0.329.

DSpace collection: 期刊論文

Double Exponential Smoothing Slime Mould Algorithm For Disease Detection In Iot Healthcare System

Role of Corporate Social Responsibility in the Financial Sustainability of Sports Organizations

Web-Based Personalized Machine Learning Recommendations to Enhance Shared Decision-Making in Prostate-Specific Antigen Screening: Randomized Controlled Trial

Decoupled Detection and Category-Level 6D Pose Estimation for Robot Grasping

Integrating deep learning and groundwater dynamics for drought vulnerability assessment under climate scenarios

A CNN-transformer framework for air quality forecasting to support aeolian dust management in river basins

A case study on the application of a data-driven (XGBoost) approach on the environmental and socio-economic perspectives of agricultural groundwater management

Advanced groundwater level forecasting with hybrid deep learning model: Tackling water challenges in Taiwan's largest alluvial fan

Intelligent Urban Flood Management Using Real-Time Forecasting, Multi-Objective Optimization, and Adaptive Pump Operation

AI-driven weather downscaling for smart agriculture using autoencoders and transformers

Flood resilience through hybrid deep learning: Advanced forecasting for Taipei's urban drainage system

Attention Distribution-Aware Softmax for NPU-Accelerated On-Device Inference of LLMs: An Edge-Oriented Approximation Design

感謝表現として多文化多言語社会の台湾にトランスフォーメーションした村上春樹の「小確幸」ランブクピティヤ ディヌーシャ編

洋行者漱石の文明開化との苦闘と回帰昇華

パートナーシップの観点から見た『ノルウェイの森』の男同士の関係変容―「死は生の対極としてではなく、その一部として存在している」の言葉に注目して―

「死の文学」と言われた村上春樹の1980年代までの創作群に築いた死生観―『ダンス・ダンス・ダンス』とその前の短篇集を中心に―

日語翻譯教育之於AI，AI之於日語翻譯教育的課題：AI賦能學習者創新自主學習的活教材

A CNN-transformer framework for air quality forecasting to support aeolian dust management in river basins

High-spatiotemporal-resolution PM2.5 forecasting by hybrid deep learning models with ensembled massive heterogeneous monitoring data

Develop a hybrid machine learning model for promoting microbe biomass production

Advancing climate-resilient flood mitigation: Utilizing transformer-LSTM for water level forecasting at pumping stations

An Artificial Intelligence-Powered Environmental Control System for Resilient and Efficient Greenhouse Farming

深度學習：環境資料數位化的應用

Watershed groundwater level multistep ahead forecasts by fusing convolutional-based autoencoder and LSTM models

Advanced groundwater level forecasting with hybrid deep learning model: Tackling water challenges in Taiwan’s largest alluvial fan

A speech enhancement method using Fast Fourier Transform and Convolutional Autoencoder

Intelligent Urban Flood Management Using Real-Time Forecasting, Multi-Objective Optimization, and Adaptive Pump Operation

Interactive urban building energy modelling with functional mockup interface of a local residential building stock

Seamless integration of convolutional and back-propagation neural networks for regional multi-step-ahead PM2.5 forecasting

Integrate deep learning and physically-based models for multi-step-ahead microclimate forecasting

Deep neural networks for spatiotemporal PM2.5 forecasts based on atmospheric chemical transport model output and monitoring data

Real-time image-based air quality estimation by deep learning neural networks

AI-driven weather downscaling for smart agriculture using autoencoders and transformers

ResNet-SE-CBAM Siamese Networks for Few-Shot and Imbalanced PCB Defect Classification

Efficient License Plate Alignment and Recognition Using FPGA‑Based Edge Computing

Multi-objective mathematical model for optimal wind turbine placement in wind farm under uncertainty

Hybrid Hierarchical Attention Network-Hierarchical Deep Learning for Text Classification in Opinion Mining

Skill Prediction and Player Re-Identification Using Serial Exponential Ring Toss Game-Based Optimization Based Deep Maxout Network

Deploying a Skeleton-Based Video Anomaly Detection System on Edge Devices for Human Activity Surveillance

A novel approach for Supply Chain Shipment Pricing Prediction using Temporal Convolutional Network- Residual Neural Network

感謝表現として多文化多言語社会の台湾にトランスフォーメーションした村上春樹の「小確幸」ランブクピティヤ　ディヌーシャ編