Multi-model Approach Research Articles

Deep learning and smart energy-based lightweight urban power load forecasting model for sustainable urban growth

IntroductionUrban power load forecasting is essential for smart grid planning but is hindered by data imbalance issues. Traditional single-model approaches fail to address this effectively, while multi-model methods mitigate imbalance by splitting datasets but incur high costs and risk losing shared power distribution characteristics.MethodsA lightweight urban power load forecasting model (DLUPLF) is proposed, enhancing LSTM networks with contrastive loss in short-term sampling, a difference compensation mechanism, and a shared feature extraction layer to reduce costs. The model adjusts predictions by learning distribution differences and employs dynamic class-center contrastive learning loss for regularization. Its performance was evaluated through parameter tuning and comparative analysis.ResultsThe DLUPLF model demonstrated improved accuracy in forecasting imbalanced datasets while reducing computational costs. It preserved shared power distribution characteristics and outperformed traditional and multi-model approaches in efficiency and prediction accuracy.DiscussionDLUPLF effectively addresses data imbalance and model complexity challenges, making it a promising solution for urban power load forecasting. Future work will focus on real-time applications and broader smart grid systems.

A Bildung theoretical framework for Models-based Practice in physical education

ABSTRACT Background: Models-based Practice (MbP) has gained popularity in the literature on physical education. It offers a comprehensive approach to teaching and programme planning. However, MbP lacks a solid theoretical foundation to guide reflections on educational purposes, teaching methods and learning aspirations. Purpose: This paper proposes a framework to address this shortcoming by drawing on Wolfgang Klafki's theory of general Bildung. The goal is to provide a theoretical foundation for MbP and contribute to its further development. By exploring the potential of a Bildung theoretical framework for MbP, this paper also seeks to bridge the gap between the Anglo-American tradition that MbP adheres to and the Continental tradition of Bildung. Relevance and Implications: The analysis demonstrates how Bildung theory can complement and challenge elements in MbP. First, it can contribute with an overarching educational purpose, which is to develop the capacity for self-determination, co-determination, and solidarity. Second, the principle of exemplarity can secure alignment between the main themes of models and epochal key problems. Third, the principles of non-affirmativity and versatility can inform learning aspirations for students, and the principle of multidimensionality can inform a multi-model approach where students engage with different body-cultural practices. Fourth, the principles of autonomy and interdependency can inform the pedagogical work of teachers. Conclusion: We conclude that, on the one hand, MbP can serve as a framework for integrating the educational principles of Bildung in schools, enriching the pedagogical landscape of physical education. On the other hand, Bildung theory can provide a solid theoretical framework for MbP, as it underpins several strengths of the approach and can guide its future development. Thus, MbP can provide a practical framework for working with general Bildung in physical education, while Bildung theory can provide a theoretical framework with criteria for the development and use of pedagogical models in MbP that align with democratic and educational objectives.

Identifying the influence of hydroclimatic factors on streamflow: A multi-model data-driven approach

Identifying the influence of hydroclimatic factors on streamflow: A multi-model data-driven approach

Age, growth, and intrinsic sensitivity of Endangered Spinetail Devil Ray (Mobula mobular) and Bentfin Devil Ray (M. thurstoni) in the Indian Ocean

AbstractDevil rays (Mobula spp.) are caught in fisheries across the Indian Ocean, with reports of significant recent declines in catch and sightings. Globally, the few populations studied have extremely low population growth rates due to low fecundity and long reproductive cycles, making them highly vulnerable to overfishing. To allow for assessment of the current sustainability of devil ray catch in the Indian Ocean, we provide estimates of age using the caudal vertebrae; somatic growth using a Bayesian, multi-model approach; maximum intrinsic rate of population increase (rmax); and fishing mortality for Endangered Spinetail Devil Ray (Mobula mobular) and Bentfin Devil Ray (M. thurstoni) sampled from small-scale fisheries catch in Indonesia, Kenya, and Pakistan. The oldest individuals of Spinetail Devil Ray (n = 79) and Bentfin Devil Ray (n = 59) were 17.5 and six years, respectively. Both species had relatively low growth coefficients (k = 0.05 and g = 0.19 year−1, respectively), with the von Bertalanffy and logistic models providing the best fitting growth models, and low rmax (0.109 and 0.107 year−1, respectively) indicating that they are highly sensitive to overexploitation. Fishing mortality F estimates (0.16 and 0.18 year−1, respectively) were higher than rmax and exploitation ratio E (0.77 and 0.80, respectively) were higher than an optimum value of 0.5 for biological sustainability for both species, suggesting that the fisheries catches of the species are unsustainable. We demonstrate an approach to assess data-poor species and apply this to two Indian Ocean devil ray species. The results highlight the urgent need for better management actions to reduce the catch of all devil rays to prevent species extinction and aid in population recovery.

Improving Heart Attack Detection through Enhanced Machine Learning and Deep Neural Networks from Multi Model Images

Accurate and timely detection of heart attacks is crucial for effective intervention and treatment. This paper presents a comprehensive study on enhancing heart attack detection using advanced machine learning (ML) and deep neural network (DNN) models, integrated through multi-model images. We propose an innovative approach that combines various machine learning techniques and deep learning architectures to improve prediction accuracy and robustness. Our methodology includes the integration of convolutional neural networks (CNNs) for feature extraction from medical imaging data, recurrent neural networks (RNNs) for analyzing time-series data, and ensemble methods for combining predictions. We systematically evaluate these models individually and in combination to determine their effectiveness in heart attack detection. Performance metrics such as accuracy, precision, recall, and F1-score are used to assess model efficacy, and comparative analyses are conducted to highlight improvements over traditional methods. The results demonstrate that the proposed multi-model approach significantly enhances prediction accuracy and reduces false positives and negatives, offering a more reliable tool for early heart attack detection. Our findings underscore the potential of integrating diverse ML and DNN techniques to address complex medical diagnosis challenges and pave the way for future research in predictive healthcare.

Fusing multi-model climate risk assessment and insurance profitability prediction: a machine learning-based cross-country comparative analysis

Climate change-induced increases in the frequency and intensity of extreme weather events year after year pose a significant challenge to the profitability of the global insurance industry. Traditional risk assessment models have limitations in predicting insurance profitability due to the difficulty in coping with the nonlinearity and complexity of climate risk. To this end, this study proposes a multi-model fusion approach that combines fuzzy assessment models, entropy weighting, linear regression, and machine learning models (LightGBM & XGBoost) to assess the impact of climate risk on the profitability of the insurance industry. By analyzing cross-country empirical data from the U.S. and U.K. insurance markets, this study reveals the differences and challenges in coping with climate risk in different countries. The findings show that climate risk significantly affects the profitability of insurance companies and that machine learning models exhibit higher accuracy and reliability in risk prediction compared to traditional methods. This paper provides an empirical basis for insurers and policymakers to address the economic impacts of climate change and makes recommendations for optimizing insurance risk management.

High prevalence of <em>Trypanosoma</em> infection in Iberian green frogs (<em>Pelophylax perezi</em>) and evidence of a negative relationship between blood parasites and two indices of frog body condition

Trypanosoma commonly parasitizes anuran hosts but very few studies have investigated ecological relationships in multiparasitized amphibians. We analysed a sample of 29 adult Iberian green frogs (Pelophylax perezi) from a monitored population in central Spain and found that 28 of these individuals (96.5%) were infected with blood parasites. The protozoa genera Lankesterella (Apicomplexa: Eimeriorina) (72.4%) and Trypanosoma (Euglenozoa: Trypanosomatida) (69%) had the highest prevalence, followed by an intraerythrocytic bacteria of the genus Aegyptianella (Pseudomonadota: Rickettsiales) (31.0%). We also report an infection by hematic microfilariae (Nematoda: Spirurida) (6.8%), which to our knowledge represents the first documented case in Iberian amphibians. Infections with more than one parasite type occurred in 62.1% of the frogs. A multimodel inference approach indicated that the infection intensities of Trypanosoma and Aegyptianella were the most important predictors, both negatively affecting the body condition of the frogs. Furthermore, the number of leeches that frogs had when captured showed a strong positive association with Trypanosoma infection intensity. This suggests that leeches act as primary vectors for Trypanosoma. Our results revealed a high taxonomic diversity of blood parasites in green frogs, thus contributing to expand our knowledge of the biodiversity of Mediterranean wetlands and highlighted the potential negative effects of certain infections on the fitness of these amphibian hosts.

EDITORIAL

Greetings to the Readers of Jurnal Teknosains!We are delighted to present the latest edition of Jurnal Teknosains, Volume 14, Number 1, December 2024. This issue highlights groundbreaking contributions in the realms of technology and science, reflecting our dedication to fostering interdisciplinary insights into the challenges and opportunities of our rapidly evolving world. One of the central themes in this edition is the intricate interaction between humans and their environment. Notably, the study on visualizing streamflow recession curves demonstrates how mathematical modeling serves as a critical tool in water resource management. Employing multimodel approaches for baseflow recession analysis, this research underscores the importance of accurate data in understanding aquifer characteristics and water storage capacities on both local and global scales. In the domain of healthcare technology, innovation in enhancing the biocompatibility of acrylic materials with natural fibers takes center stage. The incorporation of bio-based fibers, such as banana fronds, offers a sustainable and biologically safer alternative while significantly improving material performance. This promising development addresses future demands for environmentally friendly and safe materials.

Integrated Decision Support System (IDSS) for Autism Spectrum Disorder Diagnosis: A Multi-model F ramework Approach

Objectives: Development of an Integrated Decision Support System (IDSS) for Autism Spectrum Disorder (ASD) diagnosis using advanced Data Science techniques. Methods: A IDSS multi-model approach combining GBT-DL, VO-ADA, and CNN-RF were used. In evaluation, the following parameters were computed: accuracy, precision, recall, F1-score, and Kappa statistics. Findings: The results show that the GBT-DL model performed with excellent accuracy of 95.52% performance, while the VO-ADA model achieved an accuracy level of 98.60%, and the CNN-RF model has proven to have the highest accuracy at 99.15%. The IDSS demonstrates a high predictive capability across a broad age range and gives well-accounted estimates of autism severity. Multiple Data Science approaches allowed the system to surpass the existing diagnostic techniques through an aggregate ASD diagnosis. The IDSS outperformed other state-of-the-art methods based on accuracy, recall, F1-score, and Kappa statistics, which indicate its robustness and reliability. Dependence of the IDSS on contingent information reduces dependency on conventional diagnostic tests while raising the accuracy of diagnosis. Such sophisticated models result in the timely and accurate diagnosis of individuals across the autism spectrum, add substantial value to the established diagnostic methods, and are a very significant improvement over previously applied approaches. Novelty: The IDSS integrates multiple new state-of-the-art models for precise and efficient diagnosis of ASD across a vast range of ages. Keywords: Autism Spectrum Disorder (ASD), Convolutional Neural Networks-Random Forest (CNN-RF), Data Science Models, Gradient Boosted Trees-Deep Learning (GBT-DL), Integrated Decision Support System (IDSS), Vote-AdaBoost (VO-ADA)

A Comprehensive Review of Multiple Physical and Data-Driven Model Fusion Methods for Accurate Lithium-Ion Battery Inner State Factor Estimation

With the rapid global growth in demand for renewable energy, the traditional energy structure is accelerating its transition to low-carbon, clean energy. Lithium-ion batteries, due to their high energy density, long cycle life, and high efficiency, have become a core technology driving this transformation. In lithium-ion battery energy storage systems, precise state estimation, such as state of charge, state of health, and state of power, is crucial for ensuring system safety, extending battery lifespan, and improving energy efficiency. Although physics-based state estimation techniques have matured, challenges remain regarding accuracy and robustness in complex environments. With the advancement of hardware computational capabilities, data-driven algorithms are increasingly applied in battery management, and multi-model fusion approaches have emerged as a research hotspot. This paper reviews the fusion application between physics-based and data-driven models in lithium-ion battery management, critically analyzes the advantages, limitations, and applicability of fusion models, and evaluates their effectiveness in improving state estimation accuracy and robustness. Furthermore, the paper discusses future directions for improvement in computational efficiency, model adaptability, and performance under complex operating conditions, aiming to provide theoretical support and practical guidance for developing lithium-ion battery management technologies.

MULTI-MODEL APPROACH FOR MODELLING CIRCADIAN RHYTHM PHYSIOLOGICAL PARAMETERS

Present study focuses on modelling four circadian rhythmic physiological parameters (body temperature, heart rate, cortisol levels, and melatonin levels) using a multi-model approach. Four distinct models were employed: cosinor model, spline models, Fourier series (Fourier decomposition), and ARIMA models (a traditional statistical tool for modelling times series data). A convenience sample of twenty internee doctors was selected from four public sector hospitals located in Jeddah. Data were collected over a 24-hour period, ensuring the inclusion of typical circadian patterns for the four physiological parameters. The models were evaluated based on their fit and predictive accuracy, with Mean Squared Error (MSE) used as the primary metric. For extracting relevant results, R software was used and codes developed for the proposed study are given in Appendix ‘A’. Outcomes of the present study demonstrated that each model effectively captured different aspects of the circadian rhythms, with Fourier series (Fourier decomposition) and spline models providing the most flexible fits and excelled in performance accuracy. The multi-model approach enabled a comprehensive understanding of the circadian patterns, offering insights into the complex dynamics of physiological rhythms. This study highlights the importance of using diverse modelling techniques to capture the multifaceted nature of circadian rhythms, with implications for improving health outcomes through better understanding and management of biological cycles.

Improving Transferability of Physical Adversarial Attacks on Object Detectors Through Multi-Model Optimization

Physical adversarial attacks face significant challenges in achieving transferability across different object detection models, especially in real-world conditions. This is primarily due to variations in model architectures, training data, and detection strategies, which can make adversarial examples highly model-specific. This study introduces a multi-model adversarial training approach to improve the transferability of adversarial textures across diverse detection models, including one-stage, two-stage, and transformer-based architectures. Using the Truck Adversarial Camouflage Optimization (TACO) framework and a novel combination of YOLOv8n, YOLOv5m, and YOLOv3 models for optimization, our approach achieves an AP@0.5 detection score of 0.0972—over 50% lower than textures trained on single models alone. This result highlights the importance of multi-model training in enhancing attack effectiveness across object detectors, contributing to improved adversarial effectiveness.

AUTOMATED MULTI-MODEL PREDICTION AND EVALUATION FOR CONNECTING RAINFALL PREDICTION INFORMATION AND SINGLE-YEAR OPERATIONAL PLAN OF LAHOR-SUTAMI DAM

There is a gap between existing climate information and the needs of annual dam operational planning. This study aims to demonstrate that the percentile approach currently used for planning is not optimal, especially now that automation has become more accessible. The purpose of this study is to design an automated forecasting and evaluation system based on 36 10-days rainfall projections using a multi-model approach. This approach comprises a percentile, ARIMA, ECMWF+ARIMA, IOD DMI regression, ERSST regression, and ensemble methods models. Additionally, this study aims to demonstrate how a verified, multi-model-based rainfall forecast can provide more reliable assurance for the annual operational planning of Lahor-SutamiDam, simulated operationally in November 2022 for the 2022/2023 planning cycle. Data utilized include historical 10-days rainfall data from 1991 to 2023, ECMWF raw and corrected model outputs, Nino-Dipole index, and global sea surface temperature. The verification method employs four criteria based on MAE and fit index. An operational simulation approach is used for training-testing period segmentation, while a 10-year window is applied to account for possible climate-change-induced shifts in relationships. Single linear regression is used to avoid overfitting. The automation system was developed using R-Statistics. Results indicate that the current approach is only optimal for 58% of locations. Superior methods identified include ECMWFcorrected, ERSST regression, and Ensemble models. A case study for 2022/2023 demonstrates that the forecast results outperform the existing plan for at least 78% of the projected periods.

Multi-model fusion method for predicting CO2 concentration in greenhouse tomatoes

With the rapid development of greenhouse agriculture, accurate prediction of environmental parameters such as temperature, humidity, and carbon dioxide concentration is crucial for optimal crop growth. Traditional forecasting models struggle with the nonlinear and complex nature of greenhouse data, leading to challenges in model robustness. This study addresses these issues by proposing a multi-model fusion strategy for predicting CO2 concentration in greenhouse tomatoes. The proposed method integrates wavelet denoising (WT), variational mode decomposition (VMD), and long short-term memory networks (LSTM). This innovative nonlinear ensemble model effectively extracts key time series features and removes noise, while an introduced attention mechanism enhances the model’s focus on essential time steps, improving prediction accuracy. Experimental results demonstrate that the multi-model fusion approach significantly outperforms single models in terms of accuracy and stability, achieving mean absolute error (MAE) and root mean square error (RMSE) of 0.0117 and 0.0194, respectively. The proposed method offers significant advantages for CO2 prediction in greenhouse crops, providing a theoretical basis and technical support for optimizing and controlling greenhouse parameters. This contributes to the advancement of smart agriculture by offering an efficient environmental monitoring and prediction tool. Additionally, the study presents new ideas and technical solutions for addressing similar agricultural environment prediction challenges, optimizing greenhouse environment control strategies, and improving crop production efficiency.

A multi-model approach identifies ALW-II-41-27 as a promising therapy for osteoarthritis-associated inflammation and endochondral ossification

Low-grade inflammation and pathological endochondral ossification are key processes underlying the progression of osteoarthritis, the most prevalent joint disease worldwide. In this study, we employed a multi-faceted approach, integrating publicly available datasets, in silico analyses, in vitro experiments and in vivo models to identify new therapeutic candidates targeting these processes. Data mining of transcriptomic datasets identified EPHA2, a receptor tyrosine kinase associated with cancer, as being linked to both inflammation and endochondral ossification in osteoarthritis. A computational model of cellular signaling networks in chondrocytes predicted that in silico activation of EPHA2 in healthy chondrocytes increases inflammatory mediators and induces hypertrophic differentiation, a hallmark of endochondral ossification. We then evaluated the effect of EPHA2 inhibition using the tyrosine kinase inhibitor ALW-II-41-27 in cultured human chondrocytes from individuals with osteoarthritis, demonstrating significant reductions in both inflammation and hypertrophy. Additionally, systemic subcutaneous administration of ALW-II-41-27 in a mouse osteoarthritic model attenuated joint degeneration by reducing local inflammation and pathological endochondral ossification. Collectively, this study demonstrates a novel drug discovery pipeline that integrates computational, experimental, and animal models, paving the way for the development of disease-modifying treatments for osteoarthritis.

A Multi-Model Approach to Legal Judgment Prediction Using Advanced Knowledge Integration Techniques

This paper presents a multi-model approach to legal judgment prediction, emphasizing the integration of advanced knowledge techniques to enhance predictive accuracy and interpretability. As the volume of legal data continues to grow, traditional prediction methods often fall short in capturing the complexities of legal reasoning and case outcomes. By combining various modeling strategies, including rule-based systems and machine learning algorithms, this research demonstrates how a multi-model framework can leverage the strengths of different methodologies to provide more reliable predictions. Furthermore, the incorporation of knowledge integration techniques, such as knowledge graphs and Large Language Models (LLM), enriches the predictive models by offering contextual insights and improving feature selection. The findings indicate that this innovative approach not only improves the accuracy of legal predictions but also fosters transparency and trust among legal practitioners. Ultimately, this study lays the groundwork for future research in legal judgment prediction, advocating for the continued exploration of hybrid models and the application of emerging technologies to address the evolving challenges within the legal landscape.

Classification of hand movements from EEG using a FusionNet based LSTM network

Objective. Accurate classification of electroencephalogram (EEG) signals is crucial for advancing brain-computer interface (BCI) technology. However, current methods face significant challenges in classifying hand movement EEG signals, including effective spatial feature extraction, capturing temporal dependencies, and representing underlying signal dynamics.Approach. This paper introduces a novel multi-model fusion approach, FusionNet-Long Short-Term Memory (LSTM), designed to address these issues. Specifically, it integrates Convolutional Neural Networks for spatial feature extraction, Gated Recurrent Units and LSTM networks for capturing temporal dependencies, and Autoregressive (AR) models for representing signal dynamics.Main results. Compared to single models and state-of-the-art methods, this fusion approach demonstrates substantial improvements in classification accuracy. Experimental results show that the proposed model achieves an accuracy of 87.1% in cross-subject data classification and 99.1% in within-subject data classification. Additionally, Gradient Boosting Trees were employed to evaluate the significance of various EEG features to the model.Significance. This study highlights the advantages of integrating multiple models and introduces a superior classification model, which is pivotal for the advancement of BCI systems.

Assessing the Interaction Impacts of Multi-Scenario Land Use and Landscape Pattern on Water Ecosystem Services in the Greater Bay Area by Multi-Model Coupling

Water ecosystem services (WESs) are intrinsically associated with the livelihood of urban residents and are frequently disrupted by human activities. Land use and landscape patterns are key driving factors of alterations in WESs. However, existing research primarily quantifies single-factor influences and often overlooks the interactions between these factors. This study addresses this gap by employing a multi-model coupling approach, integrating the Patch-generating Land Use Simulation (PLUS), Integrated Valuation of Ecosystem Services and Trade-offs (InVEST) model, and Geographical Detector (GD) models alongside various indicators to analyse the evolution of land use, landscape patterns and WESs in the Greater Bay Area from 2000 to 2020, and to simulate spatio-temporal change patterns in different scenarios from 2030 to 2050. Additionally, this study examines the multi-factorial interactions between land use, landscape patterns, and WESs. The results indicate that (1) urbanisation steadily increased, leading to intensified landscape fragmentation, and water yield (WY) and total phosphorus (TP) consistently increased, while total nitrogen (TN) in water gradually decreased; (2) urban areas exerted the most significant impact on WY in the Greater Bay Area while Patch density (PD) had a stronger influence on WY, and Shannon’s diversity index (SHDI) had the most pronounced effect on TN and TP; (3) the interaction between any two land-use types or landscape indices exerted a greater impact on WESs compared with the impact of individual factors alone. The interaction between urban areas and cropland substantially influenced WY (q¯ = 0.634) and most strongly affected TN and TP in water (q¯ = 0.74 and 0.73, respectively). SHDI and PD had the most significant impact on WY in the economic development scenario (q¯ = 0.19) and exhibited the greatest influence on the TN and TP levels in the ecological priority scenario (q¯ = 0.12 and 0.15, respectively). Our findings can provide theoretical and technical support for the integrated scientific planning of regional water ecosystems and the development of comprehensive land use policies in the future.

Effects of asymmetric policies to achieve emissions reduction on energy trade: A North American perspective

The implementation of asymmetric emission reduction policies can not only increase the cost of reducing emissions but also reduce the effectiveness of climate policies themselves, leading to policy inefficiencies such as carbon leakage. This paper investigates the impact of asymmetric emission reduction policies on the cost-effectiveness and efficiency of climate strategies in North America. Using a model inter-comparison approach, which combines two bottom-up global models and one top-down global model, this study assesses the effects of such policies on fuel substitution, global fossil fuel trade, and emissions in North America and globally. It is the first work where a multi-model approach is used for exploring how different energy systems react to asymmetric carbon policies. This provides critical insights into regional policy design within a global emissions framework. Quantitatively, the study reveals that asymmetric carbon pricing can lead to more than 60% global emissions reduction in certain models, but can also drive trade distortions, where U.S. exemptions result in emissions rising by more than 10% compared to reference scenarios. Qualitatively, significant fuel substitution patterns across Canada, Mexico, and the U.S. demonstrate increased coal consumption when carbon prices are unevenly applied. While no global emission increase was observed, asymmetric policies result in inefficiencies between local policy costs and emissions reduction outcomes, such as rising fossil fuel trade in non-abating regions. The findings suggest that harmonising carbon policies across regions would reduce inefficiencies and minimise carbon leakage.

Approximate planning in spatial search.

How people plan is an active area of research in cognitive science, neuroscience, and artificial intelligence. However, tasks traditionally used to study planning in the laboratory tend to be constrained to artificial environments, such as Chess and bandit problems. To date there is still no agreed-on model of how people plan in realistic contexts, such as navigation and search, where values intuitively derive from interactions between perception and cognition. To address this gap and move towards a more naturalistic study of planning, we present a novel spatial Maze Search Task (MST) where the costs and rewards are physically situated as distances and locations. We used this task in two behavioral experiments to evaluate and contrast multiple distinct computational models of planning, including optimal expected utility planning, several one-step heuristics inspired by studies of information search, and a family of planners that deviate from optimal planning, in which action values are estimated by the interactions between perception and cognition. We found that people's deviations from optimal expected utility are best explained by planners with a limited horizon, however our results do not exclude the possibility that in human planning action values may be also affected by cognitive mechanisms of numerosity and probability perception. This result makes a novel theoretical contribution in showing that limited planning horizon generalizes to spatial planning, and demonstrates the value of our multi-model approach for understanding cognition.