Machine Learning Optimization Research Articles

Estimation of Dendrocalamus giganteus leaf area index by combining multi-source remote sensing data and machine learning optimization model

The Leaf Area Index (LAI) is an essential parameter that affects the exchange of energy and materials between the vegetative canopy and the surrounding environment. Estimating LAI using machine learning models with remote sensing data has become a prevalent method for large-scale LAI estimation. However, existing machine learning models have exhibited various flaws, hindering the accurate estimation of LAI. Thus, a new method for large-scale estimation of Dendrocalamus giganteus LAI was proposed, which integrates ICESat-2/ATLAS, and Sentinel-1/-2 data, and refines machine learning models through the application of Bayesian Optimization (BO), Particle Swarm Optimization (PSO), Genetic Algorithms (GA), and Simulated Annealing (SA). First, spatial interpolation was performed using the Sequential Gaussian Conditional Simulation (SGCS) method. Then, multi-source remote sensing data were leveraged to optimize feature variables through the Pearson correlation coefficient approach. Subsequently, optimization algorithms were applied to Random Forest Regression (RFR), Gradient Boosting Regression Tree (GBRT), and Support Vector Machine Regression (SVR) models, leading to efficient large-scale LAI estimation. The results showed that the BO-GBRT model achieved high accuracy in LAI estimation, with a coefficient of determination (R2) of 0.922, a root mean square error (RMSE) of 0.263, a mean absolute error (MAE) of 0.187, and an overall estimation accuracy (P1) of 92.38%. Compared to existing machine learning methods, the proposed approach demonstrated superior performance. This method holds significant potential for large-scale forest LAI inversion and can facilitate further research on other forest structure parameters.

Graphene-Based SPR Sensor Design for Bio-Alcohol Detection in the Terahertz Regime with Machine Learning Optimization Using XGBoost Regressor

Graphene-Based SPR Sensor Design for Bio-Alcohol Detection in the Terahertz Regime with Machine Learning Optimization Using XGBoost Regressor

TinyML and edge intelligence applications in cardiovascular disease: A survey.

Tiny machine learning (TinyML) and edge intelligence have emerged as pivotal paradigms for enabling machine learning on resource-constrained devices situated at the extreme edge of networks. In this paper, we explore the transformative potential of TinyML in facilitating pervasive, low-power cardiovascular monitoring and real-time analytics for patients with cardiac anomalies, leveraging wearable devices as the primary interface. To begin with, we provide an overview of TinyML software and hardware enablers, accompanied by an examination of networking solutions such as Low-power Wide area network (LPWAN) that facilitate the seamless deployment of TinyML frameworks. Following this, we delve into the methodologies of knowledge distillation, quantization, and pruning, which represent the cornerstone strategies for optimizing machine learning models to operate efficiently within resource-constrained environments. Furthermore, our discussion extends to the role of efficient deep neural networks tailored specifically for cardiovascular monitoring on wearable devices with limited computational resources. Through a comprehensive review, we analyze the applications of prominent artificial neural network architectures including Convolutional Neural Networks (CNNs), Autoencoders, Deep Belief Networks (DBNs), and Transformers in the domain of Electrocardiogram (ECG) analytics, shedding light on their efficacy and potential in advancing healthcare technology.

Seismic anisotropy prediction using ML methods: A case study on an offshore carbonate oilfield.

Estimating seismic anisotropy parameters, such as Thomson's parameters, is crucial for investigating fractured and finely layered geological media. However, many inversion methods rely on complex physical models with initial assumptions, leading to non-reproducible estimates and subjective fracture interpretation. To address these limitations, this study utilizes machine learning methods: support vector regression, extreme gradient boost, multi-layer perceptron, and a convolutional neural network. The abundance of seismic features leads to many feature combinations, making the training and testing of machine learning models challenging. Therefore, a workflow has been developed to systematically inspect seismic features and select the most appropriate one for anisotropy estimation with reasonable accuracy. Synthetic data were generated using an earth model and well data within a finite difference numerical program. After thoroughly investigating synthetic data, the amplitudes of direct and reflected waves in the time and frequency domains were selected as input features to train machine learning methods. Optimizing the machine learning hyperparameters allowed the training and testing procedures to be completed with high accuracy. Subsequently, the optimized machine learning methods were used to predict Thomsen's parameters, ε and δ, of a shaley formation in the zone area. To validate the predictions, the ε and δ estimated at a well location were compared with those obtained using a physics-based model, resulting in the least relative errors ranging from 2.92% to 7.14%.

Harnessing Statistical Analysis and Machine Learning Optimization for Heart Attack Prediction

Artificial intelligence (AI) and the decisions derived from data are pivotal arms in almost any field nowadays; in consequence, statistical analysis and machine learning (ML) have been used in biomedical research, especially in the early diagnosis of complicated disorders such as heart diseases. Early-stage prediction of cardiovascular disease can guide physicians towards better and early treatment and improved outcomes. Here, we utilize ML tools for the enhancement of clinical decision-making based on the digital metadata of the patient. The dataset comes from UCI repository and Applied methods include descriptive statistics for better understanding, correlation/covariance to tell if there is a relationship between explanatory features such as chest pain and target values, classification analysis to explore the disease and evaluate the model, Risk factor analysis to mark the most significant inputs for algorithms, clustering techniques to group patients with similar profiles, logistic regression (LR) and comparing the results after optimization using grid search method, K-means clustering, random forest (RF), ROC curve and AUC to assess the model's ability to diagnose patients. Then applying perceptron algorithm. Suitably, with simple ML algorithms, we can predict the prognosis of the disease with high accuracy

BIMSSA: enhancing cancer prediction with salp swarm optimization and ensemble machine learning approaches

BackgroundCancer rates are rising rapidly, causing global mortality. According to the World Health Organization (WHO), 9.9 million people died from cancer in 2020. Machine learning (ML) helps identify cancer early, reducing deaths. An ML-based cancer diagnostic model can use the patient’s genetic information, such as microarray data. Microarray data are high dimensional, which can degrade the performance of the ML-based models. For this, feature selection becomes essential.MethodsSwarm Optimization Algorithm (SSA), Improved Maximum Relevance and Minimum Redundancy (IMRMR), and Boruta form the basis of this work’s ML-based model BIMSSA. The BIMSSA model implements a pipelined feature selection method to effectively handle high-dimensional microarray data. Initially, Boruta and IMRMR were applied to extract relevant gene expression aspects. Then, SSA was implemented to optimize feature size. To optimize feature space, five separate machine learning classifiers, Support Vector Machine (SVM), Random Forest (RF), Extreme Learning Machine (ELM), AdaBoost, and XGBoost, were applied as the base learners. Then, majority voting was used to build an ensemble of the top three algorithms. The ensemble ML-based model BIMSSA was evaluated using microarray data from four different cancer types: Adult acute lymphoblastic leukemia and Acute myelogenous leukemia (ALL-AML), Lymphoma, Mixed-lineage leukemia (MLL), and Small round blue cell tumors (SRBCT).ResultsIn terms of accuracy, the proposed BIMSSA (Boruta + IMRMR + SSA) achieved 96.7% for ALL-AML, 96.2% for Lymphoma, 95.1% for MLL, and 97.1% for the SRBCT cancer datasets, according to the empirical evaluations.ConclusionThe results show that the proposed approach can accurately predict different forms of cancer, which is useful for both physicians and researchers.

Synthesis and characterization of novel lignocellulosic biomass-derived activated carbon for dye removal: Machine learning optimization, mechanisms, and antibacterial properties

Transforming waste materials into valuable products plays a crucial role in promoting sustainability and protecting environment. In this study, activated carbon derived from sugarcane bagasse, a lignocellulosic biomass source, was coated with iron and manganese for the adsorption and photodegradation of Acid Black 1 (AB1) dye from aqueous solutions. The effects of various parameters were investigated using Central Composite Design (CCD) and a Multi-Layer Perceptron (MLP) algorithm. The results of the characterization indicated that the iron and manganese particles were uniformly dispersed on the activated carbon. While CCD determined ideal parameters to be an initial concentration of 20.15 mg L−1, a dose of 25 mg/30 mL, a pH of 5, and a time of 40 min, the MLP Algorithm proposed slightly different conditions: an initial concentration of 26.66 mg L−1, dose of 25 mg, pH of 5.0, and a time of 25.05 min. Despite these differences, both methods projected impressive AB1 removal efficiency, 100 % for CCD and 99.02 % for MLP, underscoring the potential effectiveness of these strategies in AB1 mitigation. Concentration emerged as the predominant factor influencing the removal process, as determined by the MLP algorithm. The results showed that the major active species in the degradation of AB1 were ecb− and •O2−, while hvb + species also participated to some extent, and triethanolamine (TEOA) had a minor effect on the degradation efficiency. The nanocomposite exhibited a high antibacterial activity against Staphylococcus aureus, resulting in a large zone of inhibition. The optimized nanocomposite could be used as an effective nanomaterial to remove hazardous contaminants.

Compiler Sequence Optimization Using Machine Learning Prediction Method

Compiler optimization is crucial in improving program performance by improving execution speed, reducing memory usage, and minimizing energy consumption. Nevertheless, modern compilers, such as LLVM, with their numerous optimization passes, present a significant challenge in identifying the most effective sequence for optimizing a program. This study addresses the complex problem of determining optimal compiler optimization sequences within the LLVM framework, which encompasses 64 optimization passes, causing in an immense search space of 264264. Identifying the ideal sequence for even simple code can be an arduous task, as the interactions between passes are intricate and unpredictable. The primary objective of this research is to utilize machine-learning techniques to predict effective optimization sequences that outperform the default -O2 and -O3 optimization flags. The methodology involves generating 2,000 sequences per program and picking the one that achieves the shortest execution time. Three machine learning models—K-Nearest Neighbor (KNN), Decision Tree (DT), and Feedforward Neural Network (FFNN)—were employed to predict the optimization sequences based on features extracted from programs during execution. The study used benchmarks from Polybench, Shootout, and Stanford suites, each with varying problem sizes, to validate the proposed technique. The results demonstrate that the KNN model produced optimization sequences with superior performance compared to DT and FFNN. On average, KNN achieved execution times that were 2.5 times faster than those achieved using the O3 optimization flag. This research contributes to the field by programming the process of selecting optimal compiler sequences, which significantly reduces execution time and eliminates the need for manual tuning. It highlights the potential of machine learning in compiler optimization, offering a robust and scalable approach to improving program performance and setting the foundation for future advancements in the domain.

Analyzing failure mechanisms and predicting step-like displacement: Rainfall and RWL dynamics in lock-unlock landslides

Lock-unlock landslides have thick sliding zones that store a lot of energy. This makes them start quickly, happen suddenly, and have serious consequences. Therefore, it becomes urgent to study the deformation and failure mechanisms of such landslides and develop rational predictive models. Taking the Jiuxianping landslide as an example, this study investigates the regularity of landslide displacement changes using multi-source data, focusing on the abrupt displacement patterns in the unlock phase. Furthermore, employing Transient Release and Inhalation Method tests combined with Geo-Studio’s SEEP/W and SIGMA/W modules for fluid–solid coupled simulation calculations, the evolution process of landslide failure mechanisms and deformation characteristics is analyzed and discussed. Lastly, utilizing data mining analysis of multi-source data, a hybrid optimized machine learning predictive model is established for model prediction comparison. The study reveals that: (1) The rise in infiltration line elevates pore water pressure, affecting the stability of the sliding zone, leading to “unlock effects” and step-like displacement deformation; (2) Simulation shows that YY208 is closer to the actual situation, located at the far bank position, while YY210 is greatly influenced by the “buoyancy effect”, resulting in a slowdown in deformation velocity; (3) After data preprocessing, overall actual displacement prediction performs better than simulation displacement prediction in terms of Mean Absolute Error, Mean Squared Error and Correlation Coefficient, but noise reduction processing can improve the periodic prediction effect of simulation displacement.

Optimized Machine Learning Approaches to Combining Surface-Enhanced Raman Scattering and Infrared Data for Trace Detection of Xylazine in Illicit Opioids

Optimized Machine Learning Approaches to Combining Surface-Enhanced Raman Scattering and Infrared Data for Trace Detection of Xylazine in Illicit Opioids

Feature mining for thermoelectric materials based on interpretable machine learning.

In previous laboratory preparation processes, the selection and proportioning of specific experimental parameters often stemmed from the empirical experience of predecessors, necessitating the derivation of the optimal scheme for the target experiment through extensive trial and error. This process typically required the consumption of substantial resources and time. Thanks to the advancement of machine learning technologies, these have gradually become a powerful tool for addressing complex functional problems in material optimization. This study is based on a collected database of thermoelectric materials, aiming to clarify the correspondence between physical characteristics and the thermoelectric figure of merit, zT. The key features that can directly affect the experimental results are identified, and the features that indirectly affect the experimental results are also analyzed. By employing interpretable machine learning methods, this research analyzes critical molecular features in the characterized data, utilizing feature engineering techniques to construct and optimize machine learning models for the selected key features. Furthermore, in the subsequent model fitting and analysis phase, by comparing the efficiency of different feature combinations, the optimal feature descriptors for the current experimental system were determined. The adoption of the aforementioned improved methods endowed the related models with the potential for high-throughput screening, thereby further enhancing the efficiency of experimental optimization.

Enhancing hypertension prediction: a hybrid machine learning optimization approach

Early identification of hypertension is crucial to prevent its serious complications, which can lead to devastating health effects by threatening lifestyle quality and significantly increasing premature mortality. This study aims to evaluate the effectiveness of machine learning techniques in predicting the presence of hypertension from an unbalanced dataset consisting of 4,363 records and 35 features. To balance the dataset, we employed the synthetic minority over-sampling technique (SMOTE) algorithm. In addition, to select the most relevant features, we used ant colony optimization. Next, we applied various algorithms, including logistic regression (LR), K-nearest neighbors (KNNs), support vector machine (SVM), extra trees (ETs), and AdaBoost (AB). We also evaluated the optimization of hyperparameters using two methods: Bayesian optimization (BO) and particle swarm optimization (PSO). The results reveal that the combination of AB with BO demonstrated superior performance, with an accuracy of 97.60%, a recall of 98.93%, and a precision of 98.59%. This research emphasizes the potential of machine learning techniques for anticipating hypertension and highlights the importance of optimization techniques in improving predictive models’ performance.

Development and Management of a Decentralized Energy Model Through Machine Learning Optimization Techniques in Smart Grid

Development and Management of a Decentralized Energy Model Through Machine Learning Optimization Techniques in Smart Grid

A Building Carbon Emission Prediction Model Based on Optimized Machine Learning Model

To improve the performance of traditional machine learning model, this study employs the chaotic particle swarm optimization (CPSO) and the fuzzy cuckoo search (FCS) algorithm to enhance the BP neural network and support vector machine (SVM), leading to CPSO-BP and FCS-SVM model, respectively. Simulation results indicate that both optimization algorithms significantly enhance the predictive accuracy of their respective models. The relative error of the FCS-SVM decreases by 2.68%, while the CPSO-BP model reduces by 9.75% compared to their pre-benchmark model. Although the CPSO-BP model demonstrates a more substantial improvement in the simulation, the performance of FCS-SVM remains superior to that of CPSO-BP model. Subsequently, the FCS-SVM model was selected to forecast carbon dioxide emissions in China's construction industry from 2022 to 2025, and the results were thoroughly analyzed. The findings indicate that total carbon emissions are expected to rise over the next four years; however, the growth rate is projected to slow significantly. Additionally, carbon emission intensity is anticipated to continue declining until 2023, although the rate of decline will be minimal.

Improve carbon dioxide emission prediction in the Asia and Oceania (OECD): nature-inspired optimisation algorithms versus conventional machine learning

This paper investigates the application of three nature-inspired optimisation algorithms – SHO, MFO, and GOA – combined with four machine learning methods – Gaussian Processes, Linear Regression, MLP, and Random Forest – to enhance carbon dioxide emission prediction in the OECD – Asia and Oceania region. The study uses historical carbon dioxide emissions data, socioeconomic indicators such as GDP, population density, energy consumption, and urbanisation rates, and environmental indicators such as temperature, precipitation, and forest cover. Through comprehensive experimentation, the study evaluates the performance of each combination, revealing varying effectiveness levels. The MFO-MLP combination achieved the highest accuracy with R2 values of 0.9996 and 0.9995 and RMSE values of 11.7065 and 12.8890 for the training and testing datasets, respectively. The GOA-MLP configuration achieved R2 values of 0.9994 and 0.99934 and RMSE values of 15.01306 and 14.59333. The SHO-MLP combination, while effective, showed lower performance with R2 values of 0.9915 and 0.9946 and RMSE values of 55.4516 and 41.575. The findings suggest hybrid techniques can significantly enhance prediction accuracy compared to conventional methods. This research provides valuable insights for policymakers and stakeholders, indicating that optimised machine learning models can support more informed and effective environmental policy-making and sustainability efforts in the OECD – Asia and Oceania region. Future research should explore additional optimisation algorithms and ensemble techniques to improve prediction robustness and accuracy. These findings offer a robust tool for policymakers to forecast emissions more accurately, aiding in developing targeted strategies to reduce carbon footprints and achieve climate goals.

A Comparative Review: Biological Safety and Sustainability of Metal Nanomaterials Without and with Machine Learning Assistance

In recent years, metal nanomaterials and nanoproducts have been developed intensively, and they are now widely applied across various sectors, including energy, aerospace, agriculture, industry, and biomedicine. However, nanomaterials have been identified as potentially toxic, with the toxicity of metal nanoparticles posing significant risks to both human health and the environment. Therefore, the toxicological risk assessment of metal nanomaterials is essential to identify and mitigate potential adverse effects. This review provides a comprehensive analysis of the safety and sustainability of metallic nanoparticles (such as Au NPs, Ag NPs, etc.) in key domains such as medicine, energy, and environmental protection. Using a dual-perspective analysis approach, it highlights the unique advantages of machine learning in data processing, predictive modeling, and optimization. At the same time, it underscores the importance of traditional methods, particularly their ability to offer greater interpretability and more intuitive results in specific contexts. Finally, a comparative analysis of traditional methods and machine learning techniques for detecting the toxicity of metal nanomaterials is presented, emphasizing the key challenges that need to be addressed in future research.

Automated feature selection for early keratoconus screening optimization

In this paper, an automated feature selection (FS) method is presented to optimize machine learning (ML) models' performances, enhancing early keratoconus screening. A total of 448 parameters were analyzed from a dataset comprising 3162 observations sourced from the swept source optical coherence tomography imaging system developed by the Chinese Academy of Sciences Institute of Automation (SS-1000 CASIA OCT) and electronic health records (EHR). To identify the most relevant features, the analysis of variance (ANOVA) method was used in this study. The performance of three classifiers namely K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Artificial Neural Networks (ANN) was evaluated, yielding classification accuracies of 96.79% and 96.68% for KNN, 98.95% and 97.08% for SVM, and 95.64% and 95.62% for ANN when distinguishing between 2 and 4 keratoconus classes, respectively. The results show that selecting 50 features can significantly improve the performance of the model while reducing the computation time. The automated feature selection method can also assist ophthalmologists in better understanding the links between various ocular characteristics and keratoconus, potentially leading to advances in early diagnosis, risk prediction, and clinical management of this condition.

Optimizing machine learning models for wheat yield estimation using a comprehensive UAV dataset

Timely and accurate wheat yield forecasts using Unmanned Aircraft Vehicles (UAV) are crucial for crop management decisions, food security, and ensuring the sustainability of agriculture worldwide. While traditional machine learning algorithms have already been used in crop yield modelling, previous research used machine learning algorithms with default parameters and did not take into account the complex, non-linear relationships between model variables. Especially, the combination of vegetation indices, soil properties, solar radiation, and wheat height at the field estimation has not been deeply analysed in scientific literature. We present a machine learning based wheat yield estimation model using comprehensive UAV datasets with the implementation of hyperparameter tuning to improve model performance. The performance of the models before and after optimisations was measured using the metrics RMSE, MAE and R2, and the results showed that the models improved after tuning. Furthermore, we find that the Random Forest (RF) and Extreme Gradient Boosting (XGBoost) models outperformed other examined models. Furthermore, a non-parametric Friedman test with a Nemenyi post-hoc test indicates that the best-performing algorithms for wheat yield estimation and prediction are RF and XGBoost models. In the final step, we utilised a SHapley Additive exPlanations approach to identify the direct impact of each input variable on the yield estimation model. Among the input variables, only the Red-Edge Chlorophyll Index, the Normalised Difference Red-Edge Index and wheat height were found to be of high explanatory power in predicting wheat yield. The optimised model is 7–12% more accurate in estimating wheat yields than traditional linear models.

Digital Industrial Design Method in Architectural Design by Machine Learning Optimization: Towards Sustainable Construction Practices of Geopolymer Concrete

The construction industry’s evolution towards sustainability necessitates the adoption of environmentally friendly materials and practices. Geopolymer concrete (GeC) stands out as a promising alternative to conventional concrete due to its reduced carbon footprint and potential for cost savings. This study explores the predictive capabilities of soft computing models in estimating the compressive strength of GeC, utilizing multi-layer perceptron (MLP) neural networks and hybrid systems incorporating the Gannet Optimization Algorithm (GOA) and Grey Wolf Optimizer (GWO). A dataset comprising 63 observations from a quarry mine in Malaysia is employed, with influential parameters normalized and utilized for model development. Consequently, we integrate optimization algorithms (GOA and GWO) with MLP to fine-tune the model’s parameters and improve prediction accuracy. The models are evaluated using R2, RMSE, and VAF. Various MLP architectures are explored, evaluating transfer functions and training techniques to optimize performance. In addition, hybrid models GOA–MLP and GWO–MLP are developed, with parameters fine-tuned to enhance predictive accuracy. During the training phase, the GWO–MLP model achieved an R2 of 0.981, RMSE of 0.962, and VAF of 97.44%, compared to MLP’s R2 of 0.95, RMSE of 0.918, and VAF of 94.59%. During the testing phase, GWO–MLP also showed the best performance with an R2 of 0.976, RMSE of 1.432, and VAF of 97.51%, outperforming both MLP and GOA–MLP. The GOA–MLP model demonstrated improved performance over MLP with an R2 of 0.963, RMSE of 0.811, and VAF of 95.78% in the training phase and R2 of 0.944, RMSE of 2.249, and VAF of 92.86% in the testing phase. Hence, the results show that the GWO–MLP model consistently outperforms both MLP and GOA–MLP models. Sensitivity analysis further elucidates the impact of key parameters on compressive strength, aiding in the optimization of GeC formulations for enhanced mechanical properties. Overall, the study underscores the efficacy of machine learning models in predicting GeC compressive strength, offering insights for sustainable construction practices.

Comparing and Optimizing Four Machine Learning Approaches to Radar-Based Quantitative Precipitation Estimation

To improve radar-based quantitative precipitation estimation (QPE) methods, this study investigated the relationship between radar reflectivity (Z) and hourly rainfall intensity (R) using data from 289 precipitation events in Shanghai between September 2020 and March 2024. Two Z-R relationship models were compared in terms of their fitting performance: Z = 270.81 R1.09 (empirically fitted relationship) and Z = 300 R1.4 (standard relationship). The results show that the Z = 270.81 R1.09 model outperforms the Z = 300 R1.4 model in terms of fitting accuracy. Specifically, the Z = 270.81 R1.09 model more effectively captures the nonlinear relationship between radar reflectivity and rainfall intensity, with a higher degree of agreement between the fitted curve and the observed data points. This model demonstrated superior performance across all 289 precipitation events. This study evaluated the performance of four machine learning approaches while incorporating five meteorological features: specific differential phase shift (KDP), echo-top height (ET), vertical liquid water content (VIL), differential reflectivity (ZDR), and correlation coefficient (CC). Nine QPE models were constructed using these inputs. The key findings are as follows: (1) For models with a single-variable input, the KAN deep learning model outperformed Random Forest, Gradient Boosting Decision Trees, Support Vector Machines, and the traditional Z-R relationship. (2) When six features were used as inputs, the accuracy of the machine learning models improved significantly, with the KAN deep learning model outperforming other machine learning methods. Compared to using only radar reflectivity, the KAN deep learning model reduced the MRE by 20.78%, MAE by 4.07%, and RMSE by 12.74%, while increasing the coefficient of determination (R2) by 18.74%. (3) The integration of multiple meteorological features and machine learning optimization significantly enhanced QPE accuracy, with the KAN deep learning model performing best under varying meteorological conditions. This approach offers a promising method for improving radar-based QPE, particularly considering seasonal, weather system, and precipitation stage differentiation.