WOLFHydro: a modular framework for multi-model hydrological simulations

This survey paper explores the application of machine learning (ML) techniques in the domain of workplace injuries, focusing on three key areas: risk prediction, return to work (RTW) strategies, and demographic analysis. Through an extensive review of literature from January 2015 to July 2024, this paper examines the latest advancements in ML-driven approaches to workplace safety and identifies important research gaps. This paper highlights how classical ML techniques, such as ensemble models and decision trees, have become essential tools for identifying workplace injury risks, enabling more accurate interventions. It emphasizes the importance of leveraging ML in personalized RTW programs, which use data-driven insights to improve recovery outcomes and reduce economic demands. In the context of demographic analysis, this paper explores how ML algorithms can uncover disparities in injury rates across various age groups, industries, and occupations, underscoring the need for targeted safety measures. Moreover, research gaps are identified, particularly regarding the emerging potential of advanced ML techniques, such as deep learning and large language models (LLMs), for analyzing structured and unstructured safety data, methods that have not yet been widely applied in workplace injury research. As such, future research should apply recent advances in ML, integrating these approaches with comprehensive and accessible datasets to enhance the prediction and prevention of workplace injuries, provide more detailed analytics and insights, and improve safety protocols across all industries. This comprehensive survey is an invaluable resource for researchers and practitioners leveraging ML to address complex challenges in workplace safety.

The risk of economic failure of infill drilling campaigns increases with the increasing maturity of an oil field. The increased risk is due to the lower remaining movable oil in the late stages of a field's life compared to earlier phases. Despite the additional information owing to drilling of wells and production history, substantial geological and dynamic uncertainties remain. In this paper, an approach is presented for selecting an infill well portfolio of 10 wells from a total of 96 that explicitly accounts for these uncertainties. The geological uncertainty can be exposed by generating a multitude of geological models conditioned to known log/correlation data. These models need to be classified and a subset of representative models extracted that in turn are calibrated (history matched) to past field performance and then used for forecasting. Calibration using a suitable ensemble of representative geological models aims to maintain geological diversity while reducing the error to historical data through an objective function. Probabilistic Property Maps can then be generated to identify potential infill well locations. Probabilistic maps offer a major improvement compared to selection of infill well locations using a single geological realization. For mature assets, forecasting requires simulation of incremental oil recovery since infill wells represent incremental projects over a base case. Using an appropriate ensemble of models allows probabilistic representation of the incremental oil production and associated economics. First, individual infill well performances are forecasted. Next, the individual infill well locations are evaluated taking the history matching error into account and using utility theory to cover the risk adverseness attitude of the company. Doing so enables selection of an infill well portfolio under uncertainty and leads to a selection of well locations carrying lower risk compared to a selection of locations neglecting history match error and/or disregarding the risk attitude of a company.

We consider the correspondence problem associated with the non-rigid registration of a group of images; in particular, the theoretical basis for the derivation of the objective function that defines the 'best' correspondence across a set of images. For intra-subject registration, there is an actual physical deformation process underlying the observed deformation, but for inter-subject registration, there is no such physical process, and hence no hypothetical process that generates the observed data. This leads to the conclusion that our constructions should be based on the data alone. Such a construction is possible using criteria derived from information theory. We show how many commonly used pairwise voxel-based similarity measures can be generated using these criteria, and discuss how this approach can be extended to give a unified theoretical basis for the generation of novel objective functions in the groupwise case, where both image discrepancy and image deformation terms are included in a principled way.

Li-ion batteries are important in modern technology, especially for drones, due to their high energy density, long cycle life, and lightweight properties. Predicting their performance is crucial for enhancing drone flight safety, optimizing operations, and reducing costs. This involves using advanced techniques like machine learning (e.g., Convolutional Neural Network-CNNs, Recurrent Neural Network-RNNs), statistical modeling (e.g., Kalman Filtering), and explainable AI (e.g., SHAP, LIME, PDP) to forecast battery behavior, extend battery life, and improve drone efficiency. The study aims to develop a CNN-RNN-based ensemble model, enhanced with explainable AI, to predict key battery metrics during drone flights. The model’s predictions will aid in enhancing battery performance via continuous, data-driven monitoring, improve drone safety, optimize operations, and reduce greenhouse gas emissions through advanced recycling methods. In the present study, comparisons are made for the behaviors of two different drone Li-ion batteries, numbered 92 and 129. The ensemble model in Drone 92 showed the best performance with MAE (0.00032), RMSE (0.00067), and R2 (0.98665) scores. Similarly, the ensemble model in Drone 129 showed the best performance with MAE (0.00030), RMSE (0.00044), and R2 (0.98094) performance metrics. Similar performance results are obtained in the two predictions. However, drone 129 has a minimally lower error rate. When the Partial Dependence Plots results, which are one of the explainable AI (XAI) techniques, are interpreted with the decision tree algorithm, the effect of the Current (A) value on the model estimations in both drone flights is quite evident. When the current value is around −4, the model is more sensitive and shows more changes. This study will establish benchmarks for future research and foster advancements in drone and battery technologies through extensive testing.

Sensitivity of the SWAT model to the soil and land use data parametrisation: a case study in the Thyle catchment, Belgium

P125. Development of a novel ensemble machine learning algorithm for prediction of complications and readmission after anterior cervical spinal fusion

Mental health is an important public health concern worldwide and should be a prominent part of the healthcare industry. However, the pace of development in this field appears to be quite slow. Recently AI techniques have attracted a lot of attention in different industries, including mental health. With the help of advanced AI techniques and machine learning algorithms, a personalized care that focuses on providing emotional support catered to a specific individual has been facilitated. In this paper, we analyse various systems for mental health monitoring namely virtual counselling, precision therapy, and diagnostic systems by reviewing the algorithms and parameters used in each system. We conclude by proposing a system which combines the above-mentioned systems and is planned to provide personalized mental care.

Machine learning (ML) is also seen as an advanced technique that is only usable by highly qualified specialists. This prohibits this instrument from being utilized by many doctors and biologists in their studies. This paper’s purpose is to eradicate this obsolete perception. We claim that the recent creation of advanced high-performance ML techniques helps biomedical researchers to create competitive ML models rapidly without needing in-depth knowledge of the algorithms underlying them. This advanced system is implemented used best programming tool Python including two parts. Firstly, feature engineering and preprocessing with the Neighborhood Cleaning Rule (NCL) high-performance re-sampling procedure. Second, advanced models for high-performance machine learning, including AutoML, advanced XGBoost, and advanced ensemble bagging models. Finally, we believe that our developments would improve the way doctors interpret machine learning utilizing sophisticated and high-performance machine learning technologies and facilitate broad clinical use of Artificial Intelligence (AI) techniques.

The project, entitled “Comprehensive Health Forecasting and Care Solutions Using Predictive Analytics and AI Techniques” aims to change healthcare for the better. By using advanced artificial intelligence, it hopes to predict and manage patient health outcomes. The main goal here is to build a strong predictive model. This model will use machine learning techniques to foresee various health events. This way, we can promote proactive healthcare interventions. A key part of this project is managing patient data ethically. Ensuring privacy and sticking to legal standards is essential. The methodology involves a careful system for data collection and preprocessing. It gathers a mix of clinical and demographic datasets. Using ensemble learning models along with smart feature selection methods boosts prediction accuracy and reliability. Throughout development, the project emphasizes transparency alongside model interpretability. These aspects matter for healthcare professionals who will use these tools. The potential impact here is huge! It offers a path toward better patient outcomes and smarter allocation of healthcare resources. By giving clinicians clear and understandable predictive insights, this project aids in making informed decisions and tailoring personalized patient care. Moreover, it helps in grasping health patterns and risk factors better, encouraging innovation and research in predictive healthcare. As the project wraps up, its findings are set to greatly impact healthcare practices potentially leading to cost savings and improved patient well-being. Integrating this predictive model into clinical settings could change how healthcare providers identify and respond to health risks. This represents an important step forward in the field of healthcare analytics. In short, this project is a critical milestone in improving analytics for patients, healthcare providers, and the wider medical community alike.

With the fast growth of mobile technology, social media has become important for people to share their thoughts and feelings. Businesses and governments can make better strategic decisions when they know what the public thinks. Because of this, sentiment analysis is an important tool for figuring out how different people’s opinions are. This article presents a deeplearning ensemble model for sentiment analysis. The ensemble model proposed consists of three deep-learning models, Gated Recurrent Unit (GRU), Long Short-Term Memory (LSTM), and Bidirectional LSTM (BiLSTM), as base classifiers. AraBERT is responsible for presenting the textual input data into representative embeddings. The stacking ensemble model then captures the long-range dependencies in the embedding for a given class. As a meta-classifier, Support Vector Machine (SVM) then combines the predictions made by the stacking deep learning model. In addition, data augmentation with AraGPT was implemented to address the imbalanced dataset issues. The experimental results demonstrate that the proposed model outperforms the state-of-the-art models with an accuracy of 88.89%, 90.88%, and 88.23% on the HARD, BRAD, and Twitter datasets, respectively.

Machine Learning Techniques are gaining a lot of momentum in constant improvement of disease diagnosis. In this study, we have investigated the discriminative performance of ensemble learning model for diabetes prediction at an early stage. We have used different machine learning models and then ensemble it to improve the overall prediction accuracy. The dataset used is NHANES 2013-14 comprising of 10,172 samples and 54 feature variables for diabetes section. The feature variables used are in the form of questionnaire, a set of questions suggested by NHANES (National Health and Nutrition Examination Survey). An Ensemble model using majority voting technique was developed by combining the unweighted prediction probabilities of different machine learning models. Also, the model is evaluated and validated for real user input data for user friendliness. The overall performance was improved by Ensemble Model and had an AUC (Area under Curve) of 0.75 indicating high performance.

The recent development of high-speed data transmission over wireless networks enables multimedia broadcasting service to mobile users. Multimedia broadcasting service involves interactions among different system and network components, so it is crucial for the service provider to verify the correctness of system/service model and design, and their behaviors before a new type of service is deployed. However, due to limitations of using network simulations or scaled experimental testbeds, there has been none of research on such verification and simulation framework in 3G broadcasting networks. Therefore, we propose a simulation and analysis framework for multimedia broadcasting service over wireless networks. With concrete modeling of wireless physical channel, network, and data processing on a client device, it enables the prediction of various interesting system parameters and perceived quality of multimedia streams to users. Different models of system and network components can be plugged easily in our simulation framework for further extensions. Using this framework, we analyze the processing performance for decoding scalable videos on mobile devices in CDMA2000 wireless networks.

An ensemble model based on relevance vector machine and multi-objective salp swarm algorithm for predicting burst pressure of corroded pipelines

Solar energy prediction using linear and non-linear regularization models: A study on AMS (American Meteorological Society) 2013–14 Solar Energy Prediction Contest

The classification of objects and man-made structures on the ocean surface using synthetic aperture radar (SAR) imagery finds an important use in monitoring for icebergs and sea ice. Convolutional Neural Network (CNN) method can be employed to effectively classify objects imaged through SAR data. In this paper, the CNN performance is evaluated when three different preprocessing procedures are applied to the SAR image data to prepare the CNN's input data. Segmentation or normalization algorithms are implemented in the three procedures. Experimental results demonstrate an improvement over unsegmented and un-normalized images. Performance metrics for all three methods are approximately 94%, indicating that while some image preprocessing is required to achieve higher performance, the CNN tested is robust to the noise present in the SAR images used.