Data-driven Techniques Research Articles

Validating process mining in pediatric oncology: Demonstrating protocol adherence and variability in neuroblastoma treatment.

166 Background: Neuroblastoma is the most common extracranial tumor in children. The mortality rate for patients with metastatic disease is over 50%. In the Netherlands, 30 new patients are diagnosed annually. Current clinical treatment guidelines are complex and include multiple treatment arms consisting of chemotherapy, surgery, radiotherapy, and immunotherapy (IT). Although poor patient health and complications are known to cause adaptations to the otherwise standardized treatment regimen, the extent to which these variations occur is unknown. This study aimed to validate the use of process mining, a data-driven technique, to retrospectively examine protocol adherence and variability within current clinical guidelines based upon the GPOH-DCOG NBL2009 neuroblastoma treatment protocol. Methods: Data was extracted from electronic health records, as well as laboratory, pathology, and pharmaceutical databases, to create individual event logs for each patient. Process mining was performed using a fuzzy mining algorithm. The treatment courses for all neuroblastoma patients were mapped. Commonalities and variations were highlighted, and the throughput time for key treatment phases was calculated. Subsequently, conformance with protocol was determined. Results: Analysis of 70 patients treated between 2018-2022 showed 62 distinct treatment processes, revealing significant variations in patient care. Patients with high-risk disease showed the least variation (n=47, 42 variants). Notably, 17 variants were identified among 18 patients who died, indicating a highly personalized approach in the advanced stages of the disease. Subgroup analysis of 17 high-risk patients who received IT consisting of alternating cycles of anti-GD2 antibody and retinoic acid showed 6 variants. 12/17 patients (70.6%) completed the full IT regimen. Other treatment adjustments included shortened anti-GD2 infusions for two patients (5 and 8 days versus the prescribed 10 days, respectively), potentially indicating treatment toxicity. One patient received a 15-day infusion, exceeding the prescribed limit of 11 days. The recorded interval between the administration of anti-GD2 and retinoic acid was a median 5.5 hours (range: 60 seconds – 6.4 days), deviating from the prescribed interval of 24 hours and warranting further investigation. Conclusions: Process mining is a valuable tool for analyzing protocol performance and adherence in neuroblastoma treatment. Despite standardized treatment protocols, significant heterogeneity was observed in the treatment of neuroblastoma patients, especially those with poor prognosis. Throughput time analysis revealed variabilities in the IT regimen that warrant further investigation. This study underscores the potential of data-driven approaches to optimize treatment protocols and support evidence-based practices in clinical oncology.

Machine learning applications in sheet metal constitutive Modelling: A review

The numerical simulation of sheet metal forming processes depends on the accuracy of the constitutive model used to represent the mechanical behaviour of the materials. The formulation of these constitutive models, as well as their calibration process, has been an ongoing subject of research. In recent years, there has been a special focus on the application of data-driven techniques, namely Machine Learning, to address some of the difficulties of constitutive modelling. This review explores different methodologies for the application of Machine Learning algorithms to sheet metal constitutive modelling. These methodologies include the use of machine learning algorithms in the identification of constitutive model parameters and the replacement of the constitutive model by a metamodel created by a machine learning algorithm. A discussion about the merits and limitations of the different methodologies is presented, as well as the identification of some possible gaps in the literature that represent opportunities for future research.

Experimental validation of a low-cost maximum power point tracking technique based on artificial neural network for photovoltaic systems

Maximum power point tracking (MPPT) is a technique involved in photovoltaic (PV) systems for optimizing the output power of solar panels. Traditional solutions like perturb and observe (P&O) and Incremental Conductance (IC) are commonly utilized to follow the MPP under various environmental circumstances. However, these algorithms suffer from slow tracking speed and low dynamics under fast-changing environment conditions. To cope with these demerits, a data-driven artificial neural network (ANN) algorithm for MPPT is proposed in this paper. By leveraging the learning capabilities of the ANN, the PV operating point can be adapted to dynamic changes in solar irradiation and temperature. Consequently, it offers promising solutions for MPPT in fast-changing environments as well as overcoming the limitations of traditional MPPT techniques. In this paper, simulations verification and experimental validation of a proposed data-driven ANN-MPPT technique are presented. Additionally, the proposed technique is analyzed and compared to traditional MPPT methods. The numerical and experimental findings indicate that, of the examined MPPT methods, the proposed ANN-MPPT approach achieves the highest MPPT efficiency at 98.16% and the shortest tracking time of 1.3 s.

Empowering aged persons care services: Leveraging data-driven BiLSTM-GRU integration with wearable devices

Wearable medical devices (WMDs) have emerged as a promising medical technology to enhance patient care from a variety of medical angles. However, due to inaccurate signal measurement in sensors, the conventional model-driven framework has problems with its limited reconstruction ratio and compression ratio. A growing need for effective and individualized aged care services is created by the world's population's increasing aging. To increase the system quality and overall performance, the collected data is preprocessed to normalized data by using Z score Normalization. This study proposes a novel approach that integrates wearable device data with Elderly people's health and wellbeing may be predicted and monitored using Gated Recurrent Unit (GRU) and Bidirectional Long Short-Term Memory (BiLSTM) models. This data-driven strategy enables caregivers to deliver tailored care plans, timely treatments and early identification of potential health problems, which improve the aged population's overall quality of life. The BiLSTM-GRU evaluates the effectiveness and efficiency of an approach by analyzing the performance of Accuracy (96%), Precision (93%), F1-score (95%), Sensitivity (94%), Specificity (90%) and MSE. By advancing data-driven techniques in aged care services, this research intends to pave the way for a more effective and long-lasting healthcare system that is designed to meet the needs of the senior population.

FS/DS: A Theoretical Framework for the Dual Analysis of Feature Space and Data Space.

With the surge of data-driven analysis techniques, there is a rising demand for enhancing the exploration of large high-dimensional data by enabling interactions for the joint analysis of features (i.e., dimensions). Such a dual analysis of the feature space and data space is characterized by three components, 1) a view visualizing feature summaries, 2) a view that visualizes the data records, and 3) a bidirectional linking of both plots triggered by human interaction in one of both visualizations, e.g., Linking & Brushing. Dual analysis approaches span many domains, e.g., medicine, crime analysis, and biology. The proposed solutions encapsulate various techniques, such as feature selection or statistical analysis. However, each approach establishes a new definition of dual analysis. To address this gap, we systematically reviewed published dual analysis methods to investigate and formalize the key elements, such as the techniques used to visualize the feature space and data space, as well as the interaction between both spaces. From the information elicited during our review, we propose a unified theoretical framework for dual analysis, encompassing all existing approaches extending the field. We apply our proposed formalization describing the interactions between each component and relate them to the addressed tasks. Additionally, we categorize the existing approaches using our framework and derive future research directions to advance dual analysis by including state-of-the-art visual analysis techniques to improve data exploration.

Towards Improved Turbomachinery Measurements: A Comprehensive Analysis of Gaussian Process Modeling for a Data-Driven Bayesian Hybrid Measurement Technique

Towards Improved Turbomachinery Measurements: A Comprehensive Analysis of Gaussian Process Modeling for a Data-Driven Bayesian Hybrid Measurement Technique

Structural damage identification by using physics-guided residual neural networks

In recent years, vibration-based structural damage identification has made significant progress by exploiting data-driven deep learning techniques, which can efficiently extract damage-sensitive features from a large amount of data. However, in some practical engineering applications, large volumes of measurement data are not readily available. This paper proposes a novel physics-guided residual neural network (PhyResNet) framework to improve the robustness and accuracy of structural damage identification under data-scarce conditions. In contrast to the state-of-the-art purely data-driven ResNet, the proposed method embedded available physics knowledge (e.g., governing equations of dynamics) of structures into the feature learning process via a novel physics-based loss function. The input-output relationship of the network is constrained to retain its physical meaning implicitly while the demand for large amounts of labeled training data is reduced. Notably, even with only 5 % of the dataset used for training, PhyResNet achieves a 13.1 % improvement in R-Value. The performance of the proposed approach is evaluated through both numerical and experimental verifications. Results demonstrate that damage localization and quantification are achieved with high accuracies and good robustness.

Motor state prediction and friction compensation for brushless DC motor drives using data-driven techniques

Motor state prediction and friction compensation for brushless DC motor drives using data-driven techniques

Search for light long-lived neutral particles from Higgs boson decays via vector-boson-fusion production from pp collisions at s=13\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\sqrt{s}=13$$\\end{document} TeV with the ATLAS detector

A search is reported for long-lived dark photons with masses between 0.1 GeV and 15 GeV, from exotic decays of Higgs bosons produced via vector-boson-fusion. Events that contain displaced collimated Standard Model fermions reconstructed in the calorimeter or muon spectrometer are probed. This search uses the full LHC Run 2 (2015–2018) data sample collected in proton–proton collisions at s=13\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\sqrt{s}=13$$\\end{document} TeV, corresponding to an integrated luminosity of 139 fb-1\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$fb^{-1}$$\\end{document}. Dominant backgrounds from Standard Model processes and non-collision sources are estimated using data-driven techniques. The observed event yields in the signal regions are consistent with the expected background. Upper limits on the Higgs boson to dark photon branching fraction are reported as a function of the dark photon mean proper decay length or of the dark photon mass and the coupling between the Standard Model and the potential dark sector. This search is combined with previous ATLAS searches obtained in the gluon–gluon fusion and WH production modes. A branching fraction above 10% is excluded at 95% CL for a 125 GeV Higgs boson decaying into two dark photons for dark photon mean proper decay lengths between 173 and 1296 mm and mass of 10 GeV.

Predicting Rail Corrugation Based on Convolutional Neural Networks Using Vehicle's Acceleration Measurements.

This paper presents a deep learning approach for predicting rail corrugation based on on-board rolling-stock vertical acceleration and forward velocity measurements using One-Dimensional Convolutional Neural Networks (CNN-1D). The model's performance is examined in a 1:10 scale railway system at two different forward velocities. During both the training and test stages, the CNN-1D produced results with mean absolute percentage errors of less than 5% for both forward velocities, confirming its ability to reproduce the corrugation profile based on real-time acceleration and forward velocity measurements. Moreover, by using a Gradient-weighted Class Activation Mapping (Grad-CAM) technique, it is shown that the CNN-1D can distinguish various regions, including the transition from damaged to undamaged regions and one-sided or two-sided corrugated regions, while predicting corrugation. In summary, the results of this study reveal the potential of data-driven techniques such as CNN-1D in predicting rails' corrugation using online data from the dynamics of the rolling-stock, which can lead to more reliable and efficient maintenance and repair of railways.

Comparison of Process-Driven SWAT Model and Data-Driven Machine Learning Techniques in Simulating Streamflow: A Case Study in the Fenhe River Basin

Hydrological modeling is a crucial tool in hydrology and water resource management for analyzing runoff evolution patterns. In this study, the process-driven soil and water assessment tool (SWAT) model and data-driven machine learning techniques (XGBoost, random forest, LSTM, BILSTM, and GRU) were employed to simulate runoff at monthly and daily intervals in the Fenhe River basin, situated in the middle reaches of the Yellow River, respectively. The SWAT model demonstrated effective performance in simulating runoff at various scales, with the coefficient of determination (R2) exceeding 0.80 and the Nash–Sutcliffe efficiency (NSE) surpassing 0.79. Sensitivity analysis reveals varying degrees of sensitivity among the model parameters. Furthermore, the deep learning techniques (LSTM, BILSTM, and GRU) exhibited superior simulation generalization capabilities compared to the SWAT model across various scales. Additionally, the generalization abilities of traditional machine learning techniques (XGBoost and random forest) were comparable to the SWAT model. This indicates that deep learning techniques demonstrate remarkable stability and generalization capabilities across various scales. This analysis was motivated by the use of external continuous time series data as input and the application of deep learning techniques to internal mechanisms. Moreover, an integrated modeling approach was used to enhance simulation accuracy by combining the SWAT model with machine learning techniques. The results indicate that the integrated modeling approach improves simulation performance across various scales compared to the single-model approach. This research is significant for improving the efficiency of water resource utilization and management in the Fenhe River basin.

Adaptive Location-based Routing Protocols for Dynamic Wireless Sensor Networks in Urban Cyber-physical Systems

This study investigates the development and enhancement of adaptive location-based routing protocols within dynamic wireless sensor networks (WSNs) in urban cyber-physical systems, recommending the implementation of the study’s innovative Urban Adaptive Location-based Routing Protocol (UALRP). This innovative protocol integrates real-time data analytics and adaptive machine learning models into its algorithmic framework to dynamically optimize routing decisions based on continuously changing urban conditions. Through the utilization of data-driven simulation models and machine learning techniques, the research sought to significantly improve the efficiency, reliability, and scalability of urban WSNs. Existing protocols such as Geographic Adaptive Fidelity (GAF), Greedy Perimeter Stateless Routing (GPSR), and Dynamic Source Routing (DSR) were critically assessed under urban settings using extensive datasets detailing New York City's traffic patterns and environmental variables. The analysis demonstrated that while GPSR showed superior performance in terms of latency, throughput, and energy efficiency among the traditional protocols, the introduction of UALRP, with its advanced predictive and adaptive capabilities, can further optimize these metrics. The study affirms the critical role of enhancing location accuracy and the ongoing advancement of machine learning models within urban routing protocols. These insights advocate for the broader implementation of adaptive strategies like UALRP to foster the development of more resilient and efficient urban cyber-physical systems.

Virtual sensor-based proxy for black carbon estimation in IoT platforms

Black carbon (BC) has been under the spotlight of research during the last few years due to its non-regulation, its role in air pollution, and its hazardous effects. Given the high cost of the instrumentation needed to measure BC concentrations, data-driven techniques have been adopted to implement proxies that provide BC measurements from other sensor measurements. These sensors may present data quality issues due to maintenance actions, loss of data, or relocation, among others. In this paper, we propose a data-driven proxy model for BC estimation that is powered by a hybrid sensor array, including physical and virtual sensors created from machine learning techniques and governmental air quality monitoring networks. Therefore, the proposed method provides an accurate alternative to traditional data-driven BC proxies in scenarios where some physical sensors are unavailable. The results show how a BC proxy can be partially implemented using virtual sensors, obtaining only an increase in the estimation error of around 4%, allowing the estimation of BC levels even when some physical sensors are absent.

Blood Glucose Prediction from Nutrition Analytics in Type 1 Diabetes: A Review.

Type 1 Diabetes (T1D) affects over 9 million worldwide and necessitates meticulous self-management for blood glucose (BG) control. Utilizing BG prediction technology allows for increased BG control and a reduction in the diabetes burden caused by self-management requirements. This paper reviews BG prediction models in T1D, which include nutritional components. A systematic search, utilizing the PRISMA guidelines, identified articles focusing on BG prediction algorithms for T1D that incorporate nutritional variables. Eligible studies were screened and analyzed for model type, inclusion of additional aspects in the model, prediction horizon, patient population, inputs, and accuracy. The study categorizes 138 blood glucose prediction models into data-driven (54%), physiological (14%), and hybrid (33%) types. Prediction horizons of ≤30 min are used in 36% of models, 31-60 min in 34%, 61-90 min in 11%, 91-120 min in 10%, and >120 min in 9%. Neural networks are the most used data-driven technique (47%), and simple carbohydrate intake is commonly included in models (data-driven: 72%, physiological: 52%, hybrid: 67%). Real or free-living data are predominantly used (83%). The primary goal of blood glucose prediction in T1D is to enable informed decisions and maintain safe BG levels, considering the impact of all nutrients for meal planning and clinical relevance.

Data-driven MPPT techniques for optimizing vehicular fuel cell performance in hybrid DC microgrid

This paper aims to apply data-driven maximum power point tracking (MPPT) techniques specifically tailored for fuel cell vehicle (FCV) supported hybrid DC microgrids to enhance the power harvesting capability of fuel cell (FC) stacks. Compared to existing MPPT techniques, the current study focuses on developing and evaluating data-driven approaches for maximum power extraction by dynamically determining the operating point of FC power sources through a Zeta converter. An in-depth analysis is conducted by considering parameters such as efficiency, tracking accuracy, response time, and robustness to variations in load demand and operating conditions. The performance results validate that the developed three-layer neural network (TNN)-based MPPT gives better performance findings than Gaussian process regression (GPR), support vector regression (SVR), decision tree regression (DTR), and bagging ensemble decision tree (BEDT). In the performance evaluation phase, a vehicular FC with a rating of 1.26 kW is designed and operated within the temperature range of 320 K to 343 K for hydrogen pressure values ranging from 1 bar to 1.8 bar. For these operational conditions, the prediction accuracy value of the proposed TNN method is 99.6% while the performance values GPR, SVR, DTR, and BEDT are 99%, 98.6%, 97.2%, and 96%. In addition, system efficiency is increased by 0.98%, 1.25%, 2.51%, and 3.02% compared to GPR, SVR, DTR, and BEDT, respectively.

Generative Artificial Intelligence for Designing Multi-Scale Hydrogen Fuel Cell Catalyst Layer Nanostructures.

Multiscale design of catalyst layers (CLs) is important to advancing hydrogen electrochemical conversion devices toward commercialized deployment, which has nevertheless been greatly hampered by the complex interplay among multiscale CL components, high synthesis cost and vast design space. We lack rational design and optimization techniques that can accurately reflect the nanostructure-performance relationship and cost-effectively search the design space. Here, we fill this gap with a deep generative artificial intelligence (AI) framework, GLIDER, that integrates recent generative AI, data-driven surrogate techniques and collective intelligence to efficiently search the optimal CL nanostructures driven by their electrochemical performance. GLIDER achieves realistic multiscale CL digital generation by leveraging the dimensionality-reduction ability of quantized vector-variational autoencoder. The powerful generative capability of GLIDER allows the efficient search of the optimal design parameters for the Pt-carbon-ionomer nanostructures of CLs. We also demonstrate that GLIDER is transferable to other fuel cell electrode microstructure generation, e.g., fibrous gas diffusion layers and solid oxide fuel cell anode. GLIDER is of potential as a digital tool for the design and optimization of broad electrochemical energy devices.

Students’ behaviour analysis based on correlating thermal comfort and spatial simulations; case study of a schoolyard in Shiraz City

The significant impact of outdoor school environments on student behaviour highlights the need to study how students interact with their surroundings. This has become a central focus in recent educational research, leading scholars to use various methods to gain insight into student movement behaviour. This research introduces a novel data-driven spatial gridding technique, employing thermal comfort and space syntax evaluation through ENVI-met and depthMapX software. It validates the extracted climatic and spatial heatmaps using real-time drone footage and Python programming to map and visualize student movement patterns. The findings from the Pearson correlation analysis in the Orange Software suggest that the correlated spatial-thermal heatmap better represents students’ actual behaviour with a high correlation coefficient of 0.84 (p = 0.84). Furthermore, this study sheds light on how students’ dynamic and static behaviours are linked to the schoolyard’s thermal and spatial characteristics. In conclusion, this research innovatively combines real-time monitoring with dual-simulations to analyze student movement behaviour, providing valuable insights into specific schoolyard dynamics.

Challenges and opportunities for the application of digital twins in hard-to-abate industries: a review

The energy-intensive industry, heavily reliant on heat generation processes, faces the critical challenge of drastically reducing greenhouse gas emissions. This study reviews the potential of digital twin technology in decarbonizing combustion-dependent manufacturing processes. While primarily focusing on scholarly articles, our analysis may not cover all current industrial practices and digital twin implementations. This paper highlights challenges and opportunities in deploying practical digital twins for industrial furnaces within the Industry 4.0 framework. Digital twins, integrating real-time simulation models with predictive capabilities, aim to enhance industrial furnaces’ design, operation, and maintenance by combining high-fidelity simulations with data-driven modelling techniques. The findings indicate that digital twins can facilitate the transition to furnace electrification and adopting zero-carbon fuels, significantly reducing emissions and optimizing overall furnace performance. However, challenges like model reliability and high upfront investments in IT and connectivity infrastructures limit their widespread adoption. Despite these limitations, ongoing technological advancements predict a rapid expansion of digital twin applications in industrial furnaces, making them indispensable for achieving decarbonization goals in energy-intensive industries.

Prediction of residential building occupancy using Machine learning with integrated sensor and survey Data: Insights from a living lab in Morocco

Building occupancy information is essential for effective energy management in buildings through the adoption of energy conservation and occupant-centric control strategies. These strategies endeavor to contribute to optimizing energy consumption while ensuring occupant comfort. This study focuses on advanced occupancy modeling techniques to enhance energy efficiency in residential buildings, utilizing various data-driven techniques. Various Machine Learning models, such as Random Forest, Bayesian Network, Decision Trees, Support Vector Machines, K-Nearest Neighbors, eXtreme Gradient Boosting, and Regularized Greedy Forest, have been evaluated for predicting and classifying building occupancy. Employing a living lab approach within a residential setting, the research evaluates model performance using two ground truth datasets: IoT sensor and survey data. Experimental tests use Accuracy and F1_score as evaluation criteria, demonstrating accuracy rates ranging from 70% to 95.96%. The results highlight the performance of these models in predicting residential occupancy, offering valuable insight into two approaches to occupancy modeling. The Random Forest model performs exceptionally well in capturing occupancy trends, while the Bayesian Network model, combined with expert knowledge, provides detailed predictions of occupancy types and zones. The research also addresses challenges related to data collection, including privacy concerns. It presents effective occupancy modeling strategies for residential energy management.

A data-driven solution for intelligent power allocation of connected hybrid electric vehicles inspired by offline deep reinforcement learning in V2X scenario

The proper power allocation between multiple energy sources is crucial for hybrid electric vehicles to guarantee energy economy. As a data-driven technique, offline deep reinforcement learning (DRL) solely exploits existing data to train energy management strategy (EMS), which becomes a promising solution for intelligent power allocation. However, current offline DRL-based strategies put high demands on the quality of datasets, and it is difficult to obtain numerous high-quality samples in practice. Thus, a bootstrapping error accumulation reduction (BEAR)-based strategy is proposed to enhance the energy-saving performance with different kinds of datasets. After that, based on the advanced V2X technology, a data-driven energy management updating framework is proposed to improve both fuel economy and adaptability of EMS via multi-updating. Specifically, the framework deploys multiple V2X-based buses to collect real-time information, and updates the strategy periodically making full use of offline data. The results show that the proposed BEAR-based EMS performs better than state-of-the-art offline EMSs in terms of fuel economy, especially realizing an improvement of 2.25% when training with mixed datasets. It is also validated that the offline EMS with the updating mechanism can reduce energy costs step by step under two different kinds of initial datasets.