Machine Learning Optimization Research Articles

Optimizing ensembles machine learning, genetic algorithms, and multivariate modeling for enhanced prediction of maize yield and stress tolerance index

The frequent occurrence of drought, halting from unpredictable climate-induced weather patterns, presents significant challenges in breeding drought-tolerant maize to identify adaptable genotypes. The study explores the optimization of machine learning (ML) to predict both the grain yield and stress tolerance index (STI) of maize under normal and drought-induced stress. In total, 35 genotypes, comprising 31 hybrid candidates and four commercial varieties, were meticulously evaluated across three normal and drought-treated sites. Three popular ML were optimized using a genetic algorithm (GA) and ensemble ML to enhance data capture. Additionally, a Multi-trait Genotype-Ideotype Distance (MGIDI) was also involved to identify superior maize hybrids well-suited for drought conditions. The results highlight that the ensemble meta-models optimized by grid search exhibit robust performance with high accuracy across the testing datasets (R2 = 0.92 for grain yield and 0.82 for STI). The RF optimized by GA algorithm demonstrates slightly lower performance (R2 = 0.91 for grain yield and 0.79 for STI), surpassing the predictive performance of individual SVM-GA and KNN-GA models. Selection of the best-performing hybrids indicated that out of the six hybrids with the highest STI values, both the ensemble and MGIDI can accurately predict four hybrids, namely H06, H10, H13, and H35. Thus, combining ML with MGIDI enables researchers to discern traits for each genotype and holds promise for advancing the field of drought-tolerant maize breeding and expediting the development of resilient varieties.

Advances in machine learning optimization for classical and quantum photonics

The development and optimization of photonic devices and various other nanostructure electromagnetic devices present a computationally intensive task. Much optimization relies on finite-difference time-domain or finite element analysis simulations, which can become very computationally demanding for finely detailed structures and dramatically reduce the available optimization space. In recent years, various inverse design machine learning (ML) techniques have been successfully applied to realize previously unexplored optimization spaces for photonic and quantum photonic devices. In this review, recent results using conventional optimization methods, such as the adjoint method and particle swarm, are examined along with ML optimization using convolutional neural networks, Bayesian optimizations with deep learning, and reinforcement learning in the context of new applications to photonics and quantum photonics.

Design Optimisation of Metastructure Configuration for Lithium-Ion Battery Protection Using Machine Learning Methodology

The market for electric vehicles (EVs) has been growing in popularity, and by 2027, it is predicted that the market valuation will reach $869 billion. To support the growth of EVs in public road safety, advances in battery safety research for EV application should achieve low-cost, lightweight, and high safety protection. In this research, the development of a lightweight, crashworthy battery protection system using an excellent energy absorption capability is carried out. The lightweight structure was developed by using metastructure constructions with an arrangement of repeated lattice cellular structures. Three metastructure configurations (bi-stable, star-shaped, double-U) with their geometrical variables (thickness, inner spacing, cell stack) and material types (stainless steel, aluminium, and carbon steel) were evaluated until the maximum Specific Energy Absorptions (SEA) value was attained. The Finite Element Method (FEM) is utilised to simulate the mechanics of impact and calculate the optimum SEA of the various designs using machine learning methodology. Latin Hypercube Sampling (LHS) was used to derive the design variation by dividing the variables into 100 samples. The machine learning optimisation method utilises the Artificial Neural Networks (ANN) and Non-dominated Sorting Genetic Algorithm-II (NSGA-II) to forecast the design that produces maximum SEA. The optimum control variables are star-shaped cells consisting of one vertical unit cell using aluminium material with a cross-section thickness of 2.9 mm. The optimum design increased the SEA by 5577% compared to the baseline design. The accuracy of the machine learning prediction is also verified using numerical simulation with a 2.83% error. Four different sandwich structure configurations are then constructed using the optimal geometry for prismatic battery protection subjected to ground impact loading conditions. An optimum configuration of 6×4×1 core cells arrangement results in a maximum displacement of 7.33 mm for the prismatic battery in the ground impact simulation, which is still less than the deformation threshold for prismatic battery safety of 10.423 mm. It is shown that the lightweight metastructure is very efficient for prismatic battery protection subjected to ground impact loading conditions.

Catalyzing net-zero carbon strategies: Enhancing CO2 flux Prediction from underground coal fires using optimized machine learning models

Underground coal fires release substantial carbon dioxide (CO2), posing significant environmental and health threats. Accurate prediction of surface CO2 emissions in these areas is crucial for understanding combustion zones and contributes to the global net zero carbon strategy. Traditional data analysis methods have been inadequate for CO2 flux prediction, highlighting the necessity for advanced machine learning (ML) techniques. This study introduces four optimized ML models—General Regression Neural Networks (GRNNs) and Radial Basis Function Neural Networks (RBFNNs) coupled with Grey Wolf Optimizer (GWO), Particle Swarm Optimization (PSO), and Whale Optimization Algorithm (WOA)—to rapidly predict CO2 flux in areas affected by underground coal fires. Utilizing 223 field test samples, these models consider six key variables: soil temperature at 30 cm depth (ST-30), ambient pressure/temperature/humidity (AP/AT/AH), soil water content (SWC), and wind speed (A-WS). The results underscore the superior predictive accuracy of the GRNN model, with an RMSE of 0.074 and an R2 of 0.9995. Sensitivity analysis reveals A-WS and ST-30 as the most influential factors. Compared to traditional methods, these ML models demonstrate enhanced accuracy and efficiency, marking a significant advancement in the field. The study's findings have broader applications beyond underground coal fires, suggesting potential for these ML models in other environmental monitoring contexts, such as emissions tracking in urban areas or integration with satellite data for global environmental assessment. This methodology represents a pivotal step in environmental management and monitoring, offering scalable and adaptable solutions for various ecological challenges. By rapidly and accurately estimating CO2 flux from underground coal fires, this study contributes significantly to achieving the global net zero carbon target and sets a new benchmark in environmental ML applications.

Tree-Based Pipeline Optimization Machine Learning in Classifying Whistleblowing of Academic Misconduct

The critical issue of academic misconduct is of utmost importance in the field of education and understanding whistleblowing behaviour can be a potential measure to effectively address this issue. This paper highlights the benefits of using the Tree-based Pipeline Optimization (TPOT) framework as a user-friendly tool for implementing machine learning techniques in studying whistleblowing behaviour among students in universities in Indonesia and Malaysia. The paper demonstrates the ease of implementing TPOT, making it accessible to inexpert computing scientists, and showcases highly promising results from the whistleblowing classification models trained with TPOT. Performance metrics such as Area Under Curve (AUC) are used to measure the reliability of the TPOT framework, with some models achieving AUC values above 90%, and the best AUC was 99% by TPOT with a Genetic Programming population size of 40. The paper’s main contribution lies in the empirical demonstration and findings that resulted in achieving the optimal outcomes from the whistleblowing case study. This paper sheds light on the potential of TPOT as an easy and rapid implementation tool for AI in the field of education, addressing the challenges of academic misconduct and showcasing promising results in the context of whistleblowing classification.

Smart Resource Allocation in Mobile Cloud Next-Generation Network (NGN) Orchestration with Context-Aware Data and Machine Learning for the Cost Optimization of Microservice Applications

Mobile cloud computing (MCC) provides resources to users to handle smart mobile applications. In MCC, task scheduling is the solution for mobile users’ context-aware computation resource-rich applications. Most existing approaches have achieved a moderate service reliability rate due to a lack of instance-centric resource estimations and task offloading, a statistical NP-hard problem. The current intelligent scheduling process cannot address NP-hard problems due to traditional task offloading approaches. To address this problem, the authors design an efficient context-aware service offloading approach based on instance-centric measurements. The revised machine learning model/algorithm employs task adaptation to make decisions regarding task offloading. The proposed MCVS scheduling algorithm predicts the usage rates of individual microservices for a practical task scheduling scheme, considering mobile device time, cost, network, location, and central processing unit (CPU) power to train data. One notable feature of the microservice software architecture is its capacity to facilitate the scalability, flexibility, and independent deployment of individual components. A series of simulation results show the efficiency of the proposed technique based on offloading, CPU usage, and execution time metrics. The experimental results efficiently show the learning rate in training and testing in comparison with existing approaches, showing efficient training and task offloading phases. The proposed system has lower costs and uses less energy to offload microservices in MCC. Graphical results are presented to define the effectiveness of the proposed model. For a service arrival rate of 80%, the proposed model achieves an average 4.5% service offloading rate and 0.18% CPU usage rate compared with state-of-the-art approaches. The proposed method demonstrates efficiency in terms of cost and energy savings for microservice offloading in mobile cloud computing (MCC).

Optimizing 3D printed diamond lattice structure and investigating the influence of process parameters on their mechanical integrity using nature-inspired machine learning algorithms

This research investigates the effect of process parameters on the mechanical integrity of 3D-printed Polylactic Acid (PLA+) lattice structures, with a focus on diamond lattice configurations. Adhering to ASTM D695 standards, the study employs compression testing to analyze the impact of various factors, including layer height, infill pattern, cell size, and infill density. The printing process utilizes the Ultimaker Cura Fused Deposition Modelling method, ensuring accuracy and uniformity in sample production. Mechanical properties were assessed using a Universal Testing Machine, highlighting the influence of the studied parameters on compressive strength and specific energy absorption (SEA). Innovatively, this study integrates nature-inspired machine-learning algorithms for hyperparameter tuning. Particle Swarm Optimization (PSO) is employed to optimize the settings of Random Forest and XGBoost models, enhancing their predictive capabilities. Results indicate that XGBoost, fine-tuned with PSO, slightly outperforms its counterpart in predicting key mechanical properties. Furthermore, feature significance assessments reveal cell size as a critical factor for compressive strength. This research not only provides a comprehensive analysis of mechanical properties in 3D-printed structures but also introduces a novel approach to applying nature-inspired algorithms for machine learning optimization in additive manufacturing.

Prediction of effluent total nitrogen and energy consumption in wastewater treatment plants: Bayesian optimization machine learning methods

The control of effluent total nitrogen (TN) and total energy consumption (TEC) is a key issue in managing wastewater treatment plants. In this study, effluent TN and TEC predictive models were established by selecting influent water quality and process control indicators as input features. The prediction performance of machine learning methods under different random seeds was explored, the moving average method was used for data amplification, and the Bayesian algorithm was used for hyperparameter optimization. The results showed that compared with the traditional hyperparameter optimization method for effluent TN prediction, the coefficient of determination (R2) increased by 0.092 and 0.067, reaching 0.725, and the root mean square error (RMSE) decreased by 0.262 and 0.215 mg/L, reaching 1.673 mg/L, respectively, after Bayesian optimization and data amplification. During TEC prediction, R2 increased by 0.068 and 0.042, reaching 0.884, and the RMSE decreased by 232.444 and 197.065 kWh, reaching 1305.829 kWh, respectively.

Machine learning for multi-dimensional performance optimization and predictive modelling of nanopowder-mixed electric discharge machining (EDM)

AbstractAluminium 6061 (Al6061) is a widely used material for various industrial applications due to low density and high strength. Nevertheless, the conventional machining operations are not the best choice for the machining purposes. Therefore, amongst all the non-conventional machining operations, electric discharge machining (EDM) is opted to carry out the research due to its wide ability to cut the materials. But the high electrode wear rate (EWR) and high dimensional inaccuracy or overcut (OC) of EDM limit its usage. Consequently, nanopowder is added to the dielectric medium to address the abovementioned issues. Nanopowder mixed EDM (NPMEDM) process is a complex process in terms of performance predictability for different materials. Similarly, the interactions between the process parameters such as peak current (Ip), spark voltage (Sv), pulse on time (Pon) and powder concentration (Cp) in dielectric enhance the parametric sensitivity. In addition, the cryogenic treatment (CT) of electrodes makes the process complex limiting conventional simulation approaches for modelling inter-relationships. An alternative approach requires experimental exploration and systematic investigation to model EWR and overcutting problems of EDM. Thus, artificial neural networks (ANNs) are used for predictive modelling of the process which are integrated with multi-objective genetic algorithm (MOGA) for parametric optimization. The approach uses experimental data based on response surface methodology (RSM) design of experiments. Moreover, the process physics is thoroughly discussed with parametric effect analysis supported with evidence of microscopic images, scanning electron microscopy (SEM) and 3D surface topographic images. Based on multi-dimensional optimization results, the NT brass electrode showed an improvement of 65.02% in EWR and 59.73% in OC using deionized water. However, CT brass electrode showed 78.41% reduction in EWR and 67.79% improved dimensional accuracy in deionized water. In addition to that, CT brass electrode gave 27.69% less EWR and 81.40% improved OC in deionized water compared to kerosene oil.

An aptamer-based SERS method for rapid screening and identification of pathogens assisted by machine learning technique with robustness evaluation

Pathogen testing is one of the compulsory items for the quality control of pharmaceutical products. Up to now, rapid screening and identification of low-concentration pathogen contaminations is still full of challenge. Herein, an aptamer-coated magnetic beads (Apt-MBs)-based isolation strategy combined with culture medium turbidity observing and label-free SERS detection technique has been developed for rapid screening and detection of Escherichia coli, Salmonella Typhimurium and Staphylococcus aureus. Apt-MBs were employed for capturing and isolation of target pathogens. Then, with a simple enrichment process, the potential pathogens could be screened out by observing the turbidity of liquid culture medium. The presumptive bacteria species was furtherly identified via a silicon-based SERS chip couple with an optimized machine learning model, which was selected by directly evaluating the model’s generalization ability on test data. The results demonstrated that SoftMax classifier had the optimal performance with the recognition accuracy of 100 % for all samples. Finally, the method was successfully applied for identification of pathogens at ∼1 CFU spiked level in drug samples within 24 h, which is much shorter than that of the standard culture-based method (3–5 d). Therefore, the developed method showed a promising application in monitoring pathogen levels in pharmaceutical industries.

Identifying key soil characteristics for Francisella tularensis classification with optimized Machine learning models

Francisella tularensis (Ft) poses a significant threat to both animal and human populations, given its potential as a bioweapon. Current research on the classification of this pathogen and its relationship with soil physical–chemical characteristics often relies on traditional statistical methods. In this study, we leverage advanced machine learning models to enhance the prediction of epidemiological models for soil-based microbes. Our model employs a two-stage feature ranking process to identify crucial soil attributes and hyperparameter optimization for accurate pathogen classification using a unique soil attribute dataset. Optimization involves various classification algorithms, including Support Vector Machines (SVM), Ensemble Models (EM), and Neural Networks (NN), utilizing Bayesian and Random search techniques. Results indicate the significance of soil features such as clay, nitrogen, soluble salts, silt, organic matter, and zinc , while identifying the least significant ones as potassium, calcium, copper, sodium, iron, and phosphorus. Bayesian optimization yields the best results, achieving an accuracy of 86.5% for SVM, 81.8% for EM, and 83.8% for NN. Notably, SVM emerges as the top-performing classifier, with an accuracy of 86.5% for both Bayesian and Random Search optimizations. The insights gained from employing machine learning techniques enhance our understanding of the environmental factors influencing Ft’s persistence in soil. This, in turn, reduces the risk of false classifications, contributing to better pandemic control and mitigating socio-economic impacts on communities.

Electronic Nose and Optimized Machine Learning Algorithms for Noninfused Aroma-Based Quality Identification of Gambung Green Tea

Electronic Nose and Optimized Machine Learning Algorithms for Noninfused Aroma-Based Quality Identification of Gambung Green Tea

Advancements in machine learning for material design and process optimization in the field of additive manufacturing

Advancements in machine learning for material design and process optimization in the field of additive manufacturing

Exploring the chemical diversity of Capsicum chinense cultivars using NMR-based metabolomics and machine learning methods

The habanero pepper (Capsicum chinense) is a prominent spicy fruit integral to the historical, social, cultural, and economic fabric of the Yucatan peninsula in Mexico. This study leverages the power of 1H NMR spectroscopy coupled with machine learning algorithms to dissect the metabolomic profile of eleven C. chinense cultivars, including those grown by INIFAP (Habanero-Jaguar, Antillano-HRA 1-1, Antillano-HRA 7-1, Habanero-HAm-18A, Habanero-HC-23C, and Jolokia-NJolokia-22) and commercial hybrids (Habanero-Rey Votán, Habanero-Kabal, Balam, USAPR10117, and Rey Pakal). A total of fifty metabolites, encompassing sugars, amino acids, short-chain organic acids, and nucleosides, were identified from the 1H NMR spectra. The optimized machine learning model proficiently predicted the similarity percentage between the INIFAP-grown cultivars and commercial hybrids, thereby facilitating a comprehensive comparison. Biomarkers unique to each cultivar were delineated, revealing that the Habanero-Rey Votán cultivar is characterized by the highest concentration of sugars. In contrast, the Balam cultivar is rich in amino acids and short-chain organic acids, sharing a similar metabolomic profile with the Jolokia-NJolokia-22 cultivar. The findings of this study underscore the efficacy and reliability of NMR-based metabolomics as a robust tool for differentiating C. chinense cultivars based on their intricate chemical profiles. This approach not only contributes to the scientific understanding of the metabolomic diversity among habanero peppers but also holds potential implications for food science, agriculture, and the culinary arts.

Enhancing predictive accuracy: a comprehensive study of optimized machine learning models for ultimate load-carrying capacity prediction in SCFST columns

Enhancing predictive accuracy: a comprehensive study of optimized machine learning models for ultimate load-carrying capacity prediction in SCFST columns

Efficient Diagnosis of Autism Spectrum Disorder Using Optimized Machine Learning Models Based on Structural MRI

Autism spectrum disorder (ASD) affects approximately 1.4% of the population and imposes significant social and economic burdens. Because its etiology is unknown, effective diagnosis is challenging. Advancements in structural magnetic resonance imaging (sMRI) allow for the objective assessment of ASD by examining structural brain changes. Recently, machine learning (ML)-based diagnostic systems have emerged to expedite and enhance the diagnostic process. However, the expected success in ASD was not yet achieved. This study evaluates and compares the performance of seven optimized ML models to identify sMRI-based biomarkers for early and accurate detection of ASD in children aged 5 to 10 years. The effect of using hyperparameter tuning and feature selection techniques are investigated using two public datasets from Autism Brain Imaging Data Exchange Initiative. Furthermore, these models are tested on a local Saudi dataset to verify their generalizability. The integration of the grey wolf optimizer with a support vector machine achieved the best performance with an average accuracy of 71% (with further improvement to 71% after adding personal features) using 10-fold Cross-validation. The optimized models identified relevant biomarkers for diagnosis, lending credence to their truly generalizable nature and advancing scientific understanding of neurological changes in ASD.

Robust state of charge estimation of LiFePO4 batteries based on Sage_Husa adaptive Kalman filter and dynamic neural network

Robust and accurate state of charge estimation for LiFePO4 batteries has tremendous significance, since they are constituted the basis of reliable operation for battery management system. However, the nature flat voltage curve poses challenges in both the accuracy and robustness of state of charge estimation. Here, we present a state of charge estimation framework for LiFePO4 batteries with integrating physics calculation, adaptive Kalman filter, and machine learning, where nonlinearity of flat voltage curve can be captured with strong robustness by optimized machine learning. To address the parameter sensitivity problem for dynamic neural network, the sine cosine algorithm is adopted to obtain the optimal parameters. Then, the offline training model is established based on nonlinear autoregressive models with exogenous inputs and sine cosine algorithm, and the adaptive Kalman filter is selected for online estimation. The observed data from 26,650 and 18,650 LiFePO4 batteries at 25 and 40 °C has been used to verify the accuracy and robustness of the proposed method under four working conditions. Compared to the existing methods, the proposed method can achieve strong robustness and higher accuracy in state of charge estimation with a maximum relative error below 2 %.

Enhancing performance of multi-pressure evaporation organic Rankine Cycle/Supercritical Carbon Dioxide Brayton cycle through genetic algorithm and Machine learning optimization

The Supercritical Carbon Dioxide Brayton combined cycle is one of the most important ways to enhance the performance of the cycle. Since the considerable exergy loss resulting from large temperature differences in the heat exchangers of combined cycles, the multi-pressure evaporation concept was adopted from Organic Rankine cycle research. Two novel multi-pressure evaporation configurations, namely parallel evaporation Brayton combined cycle and series evaporation Brayton combined cycle, are designed with the purpose of enhancing SCO2 Brayton cycle performance. The performance of the two new cycles was compared with that of the baseline cycle. Moreover, Printed Circuit Heat Exchangers were employed to improve the overall compactness of the cycles. The results from model calculations were utilized to train an Artificial intelligence neural network model using machine learning techniques which allowed to simplify the cycle and improves the computational speed. Subsequently, genetic algorithm was employed to conduct multi-objective optimization on the three cycles. Finally, a detailed exploration was conducted from the perspective of thermal and exergy to elucidate the reasons for the advantages and differences between the two new cycles compared to the baseline cycle. The results indicate that the series Brayton combined cycle exhibits the highest performance, and the combined cycle of multi-pressure evaporation helps to improve the performance of the combined cycle. Compared to the baseline cycle, the combined cycle efficiency of both series and parallel combined cycles has increased from 45.54% to 46.18% and 46.29%, respectively, while the bottom cycle efficiency has improved from 3.81% to 4.93% and 5.03%. The ratio of heat exchange area to net output power increased from 0.6051 to 0.6614 and 0.6740 for the parallel and series cycles, respectively. Thermal analysis shows that, the fundamental purpose of changing the layout is to improve the specific power per unit of working medium in the cycle. Exergy analysis shows that, the two new combined cycles improve cycle efficiency by reducing exergy losses in the intermediate heat exchanger. Compared to the parallel version, the series version's staged compression allows for a lower temperature difference in intermediate heat exchanger 1, resulting in smaller exergy losses and higher cycle efficiency. This work will contribute to an enhanced understanding of Supercritical Carbon Dioxide combined cycles within the academic community and the introduction of multi-pressure evaporation will further improve the performance of these important cycles.

Machine learning optimization of compact finite volume methods on unstructured grids

Fourier analysis based optimization techniques have been developed for numerical schemes on structured grids and result in significant performance improvements. However, there lacks an effective optimization technique for numerical schemes on general unstructured grids, due to the mesh geometry complexity that makes Fourier transformation infeasible. To address this issue, an optimization framework based on machine learning is developed in this paper for numerical schemes on unstructured grids. The optimization of a compact high-order variational reconstruction on triangular grids is performed to illustrate the framework. An artificial neural network (ANN) is used to predict the optimal values of the derivative weights on cell interfaces that are the free parameters of the variational reconstruction, given the local geometry as input. The ANN is trained by minimizing the solution errors of a linear advection equation on a set of generated compact stencils, with a set of sampled sine waves as initial conditions. The developed optimization framework is applicable to general numerical schemes on unstructured grids as the training does not require explicit computation of dispersion or dissipation. Numerical results for inviscid flow problems show that the optimized variational finite volume scheme is remarkably more accurate and efficient than its non-optimized counterpart, demonstrating the effectiveness of the machine learning optimization.

Random generation of linearly constrained fuzzy measures and domain coverage performance evaluation

The random generation of fuzzy measures under complex linear constraints holds significance in various fields, including optimization solutions, machine learning, decision making, and property investigation. However, most existing random generation methods primarily focus on addressing the monotonicity and normalization conditions inherent in the construction of fuzzy measures, rather than the linear constraints that are crucial for representing special families of fuzzy measures and additional preference information. In this paper, we present two categories of methods to address the generation of linearly constrained fuzzy measures using linear programming models. These methods enable a comprehensive exploration and coverage of the entire feasible convex domain. The first category involves randomly selecting a subset and assigning measure values within the allowable range under given linear constraints. The second category utilizes convex combinations of constrained extreme fuzzy measures and vertex fuzzy measures. Then we employ some indices of fuzzy measures, objective functions, and distances to domain boundaries to evaluate the coverage performance of these methods across the entire feasible domain. We further provide enhancement techniques to improve the coverage ratios. Finally, we discuss and demonstrate potential applications of these generation methods in practical scenarios.