Traditional Machine Learning Models Research Articles

A Study on Enhancing Crop Yield Prediction with Accuracy through Machine and Deep Learning Models

Accurate crop yield prediction is essential for optimizing agricultural resource allocation and supporting farmers in making informed crop management decisions. This study investigates the effectiveness of various machine learning and deep learning models for predicting crop yields based on key agricultural parameters, including Nitrogen (N), Phosphorous (P), Potassium (K), temperature, humidity, pH, and rainfall. We apply multiple algorithms—Logistic Regression, Multilayer Perceptron, Sequential Minimal Optimization (SMO), J48, Random Forest, REP Tree, and a proposed deep learning model—to evaluate their predictive accuracy in classifying agricultural items. Each model’s performance is assessed using metrics such as Kappa statistic, Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), Relative Absolute Error (RAE), Root Relative Squared Error (RRSE), and the time taken to build the model. Findings reveal that while traditional machine learning models like Random Forest and REP Tree show robust performance, the proposed deep learning model demonstrates enhanced accuracy and efficiency, making it highly promising for crop yield prediction applications. This study underscores the potential of advanced learning algorithms to improve agricultural productivity by providing actionable insights for crop planning and resource management.

Conformal novelty detection for multiple metabolic networks.

Graphical representations are useful to model complex data in general and biological interactions in particular. Our main motivation is the comparison of metabolic networks in the wider context of developing noninvasive accurate diagnostic tools. However, comparison and classification of graphs is still extremely challenging, although a number of highly efficient methods such as graph neural networks were developed in the recent decade. Important aspects are still lacking in graph classification: interpretability and guarantees on classification quality, i.e., control of the risk level or false discovery rate control. In our contribution, we introduce a statistically sound approach to control the false discovery rate in a classification task for graphs in a semi-supervised setting. Our procedure identifies novelties in a dataset, where a graph is considered to be a novelty when its topology is significantly different from those in the reference class. It is noteworthy that the procedure is a conformal prediction approach, which does not make any distributional assumptions on the data and that can be seen as a wrapper around traditional machine learning models, so that it takes full advantage of existing methods. The performance of the proposed method is assessed on several standard benchmarks. It is also adapted and applied to the difficult task of classifying metabolic networks, where each graph is a representation of all metabolic reactions of a bacterium and to real task from a cancer data repository. Our approach efficiently controls - in highly complex data - the false discovery rate, while maximizing the true discovery rate to get the most reasonable predictive performance. This contribution is focused on confident classification of complex data, what can be further used to explore complex human pathologies and their mechanisms.

Causality-Driven Feature Selection for Calibrating Low-Cost Airborne Particulate Sensors Using Machine Learning

With escalating global environmental challenges and worsening air quality, there is an urgent need for enhanced environmental monitoring capabilities. Low-cost sensor networks are emerging as a vital solution, enabling widespread and affordable deployment at fine spatial resolutions. In this context, machine learning for the calibration of low-cost sensors is particularly valuable. However, traditional machine learning models often lack interpretability and generalizability when applied to complex, dynamic environmental data. To address this, we propose a causal feature selection approach based on convergent cross mapping within the machine learning pipeline to build more robustly calibrated sensor networks. This approach is applied in the calibration of a low-cost optical particle counter OPC-N3, effectively reproducing the measurements of PM1 and PM2.5 as recorded by research-grade spectrometers. We evaluated the predictive performance and generalizability of these causally optimized models, observing improvements in both while reducing the number of input features, thus adhering to the Occam’s razor principle. For the PM1 calibration model, the proposed feature selection reduced the mean squared error on the test set by 43.2% compared to the model with all input features, while the SHAP value-based selection only achieved a reduction of 29.6%. Similarly, for the PM2.5 model, the proposed feature selection led to a 33.2% reduction in the mean squared error, outperforming the 30.2% reduction achieved by the SHAP value-based selection. By integrating sensors with advanced machine learning techniques, this approach advances urban air quality monitoring, fostering a deeper scientific understanding of microenvironments. Beyond the current test cases, this feature selection method holds potential for broader applications in other environmental monitoring applications, contributing to the development of interpretable and robust environmental models.

Optoelectronic performance prediction of HgCdTe homojunction photodetector in long wave infrared spectral region using traditional simulations and machine learning models.

This research explores the design of an infrared (IR) photodetector using mercury cadmium telluride (Hg1-xCdxTe). It proposes two- and three-dimensional homojunction models based on p+-Hg0.7783Cd0.2217Te/n--Hg0.7783Cd0.2217Te, focusing on applications in the long-wavelength infrared range. The photodetector's performance is analyzed using Silvaco ATLAS TCAD software and compared with analytical calculations based on drift-diffusion, tunneling, and Chu's approximation techniques. Optimized for operation at 10.6μm wavelength under liquid nitrogen temperature, the proposed photodetector demonstrates promising optoelectronic characteristics including the dark current density of 0.20mA/cm2, photocurrent density of 4.98A/cm2, and photocurrent density-to-dark current density ratio of 2.46 × 104, a 3-dB cut-off frequency of 104GHz, a rise time of 0.8 ps, quantum efficiency of 58.30%, peak photocurrent responsivity of 4.98A/W, specific detectivity of 3.96 × 1011cmHz1/2/W, and noise equivalent power of 2.52 × 10-16W/Hz1/2 indicating its potential for low-noise, high-frequency and fast-switching applications. The study also incorporates machine learning regression models to validate simulation results and provide a predictive framework for performance optimization, evaluating these models using various statistical metrics. This comprehensive approach demonstrates the synergy between advanced materials science and computational techniques in developing next-generation optoelectronic devices. By combining theoretical modeling, simulation, and machine learning, the research highlights the potential to accelerate progress in IR detection technology and enhance device performance and efficiency. This multidisciplinary methodology could serve as a model for future studies in optoelectronics, illustrating how advanced materials and computational methods can be utilized to enhance device capabilities.

Robust Human Activity Recognition for Intelligent Transportation Systems Using Smartphone Sensors: A Position-Independent Approach

This study explores Human Activity Recognition (HAR) using smartphone sensors to address the challenges posed by position-dependent datasets. We propose a position-independent system that leverages data from accelerometers, gyroscopes, linear accelerometers, and gravity sensors collected from smartphones placed either on the chest or in the left/right leg pocket. The performance of traditional machine learning algorithms (Decision Trees (DT), K-Nearest Neighbors (KNN), Random Forest (RF), Support Vector Classifier (SVC), and XGBoost) is compared against deep learning models (Gated Recurrent Unit (GRU), Long Short-Term Memory (LSTM), Temporal Convolutional Networks (TCN), and Transformer models) under two sensor configurations. Our findings highlight that the Temporal Convolutional Network (TCN) model consistently outperforms other models, particularly in the four-sensor non-overlapping configuration, achieving the highest accuracy of 97.70%. Deep learning models such as LSTM, GRU, and Transformer also demonstrate strong performance, showcasing their effectiveness in capturing temporal dependencies in HAR tasks. Traditional machine learning models, including RF and XGBoost, provide reasonable performance but do not match the accuracy of deep learning models. Additionally, incorporating data from linear accelerometers and gravity sensors led to slight improvements over using accelerometer and gyroscope data alone. This research enhances the recognition of passenger behaviors for intelligent transportation systems, contributing to more efficient congestion management and emergency response strategies.

Explainable Machine Learning Models for Real-Time Threat Detection in Cybersecurity

In the rapidly evolving landscape of cybersecurity, traditional machine learning models often operate as "black boxes," providing high accuracy but lacking transparency in decision-making. This lack of explainability poses challenges for trust and accountability, especially in critical areas like threat detection and incident response. Explainable machine learning models aim to address this by making the model's predictions more understandable and interpretable to users. This research integrates explainable machine learning models for real-time threat detection in cybersecurity. Data from multiple sources, including network traffic, system logs, and user behavior, undergo preprocessing such as cleaning, feature extraction, and normalization. The processed data is passed through various machine learning models, including traditional approaches like SVM and decision trees, as well as deep learning models like CNN and RNN. Explainability techniques such as LIME, SHAP, and attention mechanisms provide transparency, ensuring interpretable predictions. The explanations are delivered through a user interface that generates alerts, visualizations, and reports, facilitating effective threat assessment and incident response in decision support systems. This framework enhances model performance, trust, and reliability in complex cybersecurity scenarios.

Human in Loop Machine Learning: A Paradigm Shift

This paper explores the transformative concept of Human-in-Loop (HIL) Machine Learning, which integrates human expertise into the machine learning process to enhance data quality, model accuracy, and ethical decision-making. Human interaction is added to traditional machine learning steps such data collection, preprocessing, model training, evaluation, and deployment through continuous feedback loops, data annotation, and model change. This integration makes use of human intuition and experience to enhance algorithm performance and model interpretability, hence mitigating the drawbacks of entirely automated approaches. HIL Machine Learning allows AI systems to adjust to changing obstacles and makes more accurate forecasts by promoting human-machine interaction. This study emphasizes HIL Machine Learning as a reliable method for successfully handling dynamic, complicated problems across a range of areas.

Prognostic importance of the Scottish inflammatory prognostic score in patients with hepatocellular carcinoma after hepatectomy: a retrospective cohort study.

The Scottish Inflammatory Prognostic Score (SIPS), an innovative scoring system, has emerged as a promising biomarker for predicting patient outcomes following cancer therapy. This study aimed to evaluate the value of SIPS as a prognostic indicator following hepatectomy in patients with hepatocellular carcinoma (HCC). This retrospective study included 693 HCC patients who underwent hepatectomy. Survival outcomes were compared between propensity score-matched groups. Independent prognostic factors were identified through Cox regression analysis. Additionally, both traditional Cox proportional hazards models and machine learning models based on the SIPS were developed and validated. A total of 693 HCC patients who underwent hepatectomy were included, with 102 in the high SIPS group and 591 in the low SIPS group. Following propensity score matching (1:3 ratio), both groups achieved balance, with 82 patients in the high SIPS group and 240 patients in the low SIPS group. The low SIPS group demonstrated significantly superior recurrence-free survival (RFS) (25 months vs. 21 months; P < 0.001) and overall survival (OS) (69 months vs. 58 months; P < 0.001) compared to the high SIPS group. Multivariable analysis identified SIPS as an independent adverse factor affecting both RFS and OS. The calibration curve for overall patient survival diagnosis displayed excellent predictive accuracy. Traditional COX prognostic models and machine learning models incorporating SIPS demonstrated excellent performance both the training and validation set. This study confirms the prognostic significance of SIPS in post-hepatectomy HCC patients, providing a practical tool for risk stratification and clinical decision-making. Further research and validation are needed to consolidate its role in prognostic assessment.

UAV Detection using Convolutional Neural Networks and Radio Frequency Data

Unmanned Aerial Vehicles (UAVs) are increasingly seen as significant threats to both civilian and military installations, highlighting the need for effective UAV detection systems. Among the various detection methods proposed, passive radio frequency sensing has proven particularly effective. In this study, we utilized an open radio frequency dataset to explore the use of Convolutional Neural Networks (CNNs) for UAV detection. Our comparison with traditional machine learning models, such as random forest, AdaBoost, and XGBoost, demonstrates that CNNs offer superior performance. Specifically, CNNs achieved impressive accuracy rates: 99.93% for detecting UAVs, 93.30% for classifying UAV types, and 76.43% for categorizing UAV modes. These results underscore the potential of CNNs as a powerful tool for UAV detection and classification.

A deep learning model for prediction of autism status using whole-exome sequencing data.

Autism is a developmental disability. Research demonstrated that children with autism benefit from early diagnosis and early intervention. Genetic factors are considered major contributors to the development of autism. Machine learning (ML), including deep learning (DL), has been evaluated in phenotype prediction, but this method has been limited in its application to autism. We developed a DL model, the Separate Translated Autism Research Neural Network (STAR-NN) model to predict autism status. The model was trained and tested using whole exome sequencing data from 43,203 individuals (16,809 individuals with autism and 26,394 non-autistic controls). Polygenic scores from common variants and the aggregated count of rare variants on genes were used as input. In STAR-NN, protein truncating variants, possibly damaging missense variants and mild effect missense variants on the same gene were separated at the input level and merged to one gene node. In this way, rare variants with different level of pathogenic effects were treated separately. We further validated the performance of STAR-NN using an independent dataset, including 13,827 individuals with autism and 14,052 non-autistic controls. STAR-NN achieved a modest ROC-AUC of 0.7319 on the testing dataset and 0.7302 on the independent dataset. STAR-NN outperformed other traditional ML models. Gene Ontology analysis on the selected gene features showed an enrichment for potentially informative pathways including potassium ion transport.

Pneumonia detection on chest X-rays from Xception-based transfer learning and logistic regression.

Pneumonia is a dangerous disease that kills millions of children and elderly patients worldwide every year. The detection of pneumonia from a chest x-ray is perpetrated by expert radiologists. The chest x-ray is cheaper and is most often used to diagnose pneumonia. However, chest x-ray-based diagnosis requires expert radiologists which is time-consuming and laborious. Moreover, COVID-19 and pneumonia have similar symptoms which leads to false positives. Machine learning-based solutions have been proposed for the automatic prediction of pneumonia from chest X-rays, however, such approaches lack robustness and high accuracy due to data imbalance and generalization errors. This study focuses on elevating the performance of machine learning models by dealing with data imbalanced problems using data augmentation. Contrary to traditional machine learning models that required hand-crafted features, this study uses transfer learning for automatic feature extraction using Xception and VGG-16 to train classifiers like support vector machine, logistic regression, K nearest neighbor, stochastic gradient descent, extra tree classifier, and gradient boosting machine. Experiments involve the use of hand-crafted features, as well as, transfer learning-based feature extraction for pneumonia detection. Performance comparison using Xception and VGG-16 features suggest that transfer learning-based features tend to show better performance than hand-crafted features and an accuracy of 99.23% can be obtained for pneumonia using chest X-rays.

Multi-classification prediction of PM2.5 concentration based on improved adaptive boosting rotation forest

Developing efficient, stable, and user-friendly methods and technologies for predicting air quality has contributed to environmental research and management. Most traditional machine learning (ML) models often struggle to efficiently process extensive air quality data and grapple with the challenge of imbalanced data distributions. To this end, we introduced a novel multi-strategy collaborative approach that incorporates weighted feature selection, an adaptive enhanced rotation forest algorithm, and Bayesian Optimization for parameter tuning. Moreover, to improve the transparency in black box ML models, the novel Shapley Additive Explanations (SHAP) method was applied to interpret the outputs and analyze the importance of individual variables. Through experiments on real air quality datasets, we have verified the accuracy of our proposed method in classifying air quality levels. Notably, our model demonstrated an average test set accuracy improvement of approximately 10 % in cities like Beijing, Tianjin, Shijiazhuang, and Baoding after hyperparameter optimization using Bayesian methods. Furthermore, compared to alternative algorithms, our model showed an improvement of 1–2 % in evaluation metrics. By introducing novel methodologies and tools, our research contributes significantly to the advancement of air quality classification technology and holds profound implications for informing environmental policy decisions.

Uncovering suggestions in MOOC discussion forums: a transformer-based approach

The field of natural language processing has experienced significant advances in recent years, but these advances have not yet resulted in improved analytics for instructors on MOOC platforms. Valuable information, such as suggestions, is generated in the comment forums of these courses, but due to their volume, manual processing is often impractical. This study examines the feasibility of fine-tuning and effectively utilizing state-of-the-art deep learning models to identify comments that contain suggestions in MOOC forums. The main challenges encountered are the lack of labeled datasets from the MOOC context for fine-tuning classification models and the soaring computational cost of this training. For this study, we manually collected and labeled 2228 comments in Spanish and English from 5 MOOCs and scraped 1.4 million MOOC reviews from 3 platforms. We fine-tuned and evaluated 4 pretrained models based on the transformer architecture and 3 traditional machine learning models to compare their effectiveness in the suggestion mining task in this domain. Transformer-based models proved to be highly effective in this task/domain combination, achieving performance levels that matched or exceeded those deemed appropriate in other contexts and were significantly greater than those achieved by traditional models. Domain adaptation led to improved linguistic understanding of the target domain; however, in this project, this approach did not translate into an observable improvement in suggestion mining. The automated identification of comments that can be labeled as suggestions can result in considerable time savings for instructors, especially considering that less than a quarter of the analyzed comments contain suggestions.

Time series-based machine learning for forecasting multivariate water quality in full-scale drinking water treatment with various reagent dosages

Accurately predicting drinking water quality is critical for intelligent water supply management and for maintaining the stability and efficiency of water treatment processes. This study presents an optimized time series machine learning approach for accurately predicting multivariate drinking water quality, explicitly considering the time-dependent effects of reagent dosing. By leveraging data from a full-scale treatment plant, we constructed feature-engineered time series datasets incorporating influent water quality parameters, reagent dosages and effluent water characteristics. Seven predictive models, including both traditional machine learning (ML) and deep learning (DL) models were developed and rigorously evaluated against a naive mean baseline model. Our results demonstrate that traditional ML models, enhanced with time feature engineering, rivaled the performance of both widely used DL models and the naive mean baseline model. Specifically, an XGBoost model achieved superior prediction accuracy in dynamically forecasting four water quality characteristics at a 12-hour time lag step, outperforming the naive baseline model by 3-4% in terms of Mean Absolute Percentage Error (MAPE). This finding underscores the importance of incorporating a 12-hour interval to effectively capture the delayed impact of reagent dosing on water quality prediction. Furthermore, SHAP model interpretability analysis provided valuable insights into the XGBoost model's decision-making process, revealing its strong data-driven foundation aligned with established water treatment principles. This research highlights the significant potential of optimized machine learning techniques for enhancing water purification processes and enabling more informed, data-driven decision-making in the water supply industry.

Spatiotemporal estimation of surface NO2 concentrations in the Pearl River Delta region based on TROPOMI data and machine learning

Nitrogen dioxide (NO2) is a major air pollutant, and its concentration data are crucial for the study of air pollution and its impact on the environment. Although satellite data provide an effective method for estimating surface concentrations on a large scale through integrated modeling, the estimation of surface NO2 concentrations is hampered by the substantial amount of missing satellite data. This restricts in-depth studies of surface NO2 pollution. This study aims to reconstruct the missing data on tropospheric NO2 vertical column density from the TROPOspheric Monitoring Instrument (TROPOMI NO2). Subsequently, the reconstructed TROPOMI NO2 data and other predictor variables were utilized to estimate the daily surface NO2 concentrations at a 1 km resolution for the Pearl River Delta (PRD) region. The TROPOMI NO2 reconstruction models and the surface NO2 estimation model were both developed using the XGBoost algorithm. Additionally, comparative experiments were conducted between the XGBoost model and other traditional machine learning models, and the performances of the XGBoost model were evaluated through 10-fold cross-validation (CV) sample-based and site-based evaluations. The results indicate that the sample-based and site-based CV R2 values were 0.873 and 0.709, respectively. The feature importance scores indicate that TROPOMI NO2 was the most significant variable contributing to the estimation model. This indicates that the reconstruction of TROPOMI NO2 data and the development of an Extreme Gradient Boosting (XGBoost) model are suitable for the spatiotemporal estimation of surface NO2 concentrations in the PRD region, effectively reflecting the spatiotemporal distribution and evolution of surface NO2 concentrations in the area.

Hyperspectral Inversion of Soil Cu Content in Agricultural Land Based on Continuous Wavelet Transform and Stacking Ensemble Learning

Heavy metal pollution in agricultural land poses significant threats to both the ecological environment and human health. Therefore, the rapid and accurate prediction of heavy metal content in agricultural soil is crucial for environmental protection and soil remediation. Acknowledging the limitations of traditional single linear or nonlinear machine learning models in terms of prediction accuracy, this study developed an ensemble learning model that integrates multiple linear or nonlinear learning models with a random forest (RF) model to improve both the prediction accuracy and reliability. In this study, we selected a typical copper (Cu) polluted area in the Pearl River Delta of Guangdong Province as the research site and collected Cu content data and indoor soil reflectance spectral data from 269 surface soil samples. First, the soil spectral data were preprocessed using Savitzky–Golay (SG) smoothing, multiplicative scattering correction (MSC), and continuous wavelet transform (CWT) to reduce noise interference. Next, principal components analysis (PCA) was employed to reduce the dimensionality of the preprocessed spectral data, eliminating redundant features and lowering the computational complexity. Finally, based on the dimensionality-reduced data and Cu content, we established a stacked ensemble learning model, where the base models included SVR, PLSR, BPNN, and XGBoost, with RF serving as the meta-model to estimate the soil heavy metal content. To evaluate the performance of the stacking model, we compared its prediction accuracy with that of individual models. The results indicate that, compared to the traditional machine learning models, the prediction accuracy of the stacking model was superior (R2 = 0.77; RMSE = 7.65 mg/kg; RPD = 2.29). This suggests that the integrated algorithm demonstrates a greater robustness and generalization capability. This study presents a method to improve soil heavy metal content estimation using hyperspectral technology, ensuring a robust model that supports policymakers in making informed decisions about land use, agriculture, and environmental protection.

Machine learning powered predictive modelling of complex residual stress for nuclear fusion reactor design

Fusion In-vessel components, assembled and maintained using laser welding, one of the most promising techniques, exhibit complex distributions of residual stress, microstructures, and material properties. These residual stresses can compromise structural integrity and lifespan of critical components. Although using advanced experimental measurements can evaluate the residual stress for individual case, extending the measurements to massive number of components are costly and time-consuming. Traditional machine learning (ML) models struggle to account for the heterogeneity and anisotropy of these stress distributions. Here, we develop a novel ML framework based on the Eurofer97 steel, the structural material for in-vessel components. The ML framework is trained on high-resolution residual stress data derived from recently-developed evaluation techniques. Combining with microstructures, the model enables prediction of heterogenous and anisotropic residual stress distribution. It successfully predicts the compressive residual stress in fusion zone (∼−200 MPa) balanced by tensile residual stress in heat affected zone (∼300 MPa), aligning closely with experimental results with the R-squared value of 0.989 and the mean square error of 10−4. Unlike experiments that take hours, the ML model provides predictions within seconds. It offers valuable insights into residual stress prediction for various joints, enhancing the reliability and lifetime prediction of in-vessel components.

Enhancing water quality prediction with advanced machine learning techniques: An extreme gradient boosting model based on long short-term memory and autoencoder

Achieving carbon neutrality in the pulp and paper industry necessitates effectively recycling pulp and papermaking wastewater, where continuous monitoring of effluent quality indices is crucial. This study suggests a novel machine learning-based model named LSTMAE-XGBOOST that integrates the feature extraction capabilities of autoencoder, the sequential feature learning capabilities of LSTM, and the high prediction accuracy of XGBOOST. This model is capable of extracting the complex relationships, non-Gaussian characteristics, and time series features from the papermaking wastewater data, and it demonstrates superior predictive performance. Compared to traditional machine learning models, the proposed model exhibits higher prediction accuracy. Specifically, when contrasted with partial least squares regression, LSTMAE-XGBOOST achieves a 40% increase in R2 and a 35% reduction in RMSE. Further comparative assessments against other machine learning-based hybrid models with similar structures confirm the superiority of integrating LSTM and XGBOOST within the hybrid model approach. This study contributes a compelling methodology for modeling effluent quality indices, offering significant implications for environmental management in the pulp and paper industry.

Enhancing early detection of cognitive decline in the elderly: a comparative study utilizing large language models in clinical notes

BackgroundLarge language models (LLMs) have shown promising performance in various healthcare domains, but their effectiveness in identifying specific clinical conditions in real medical records is less explored. This study evaluates LLMs for detecting signs of cognitive decline in real electronic health record (EHR) clinical notes, comparing their error profiles with traditional models. The insights gained will inform strategies for performance enhancement. MethodsThis study, conducted at Mass General Brigham in Boston, MA, analysed clinical notes from the four years prior to a 2019 diagnosis of mild cognitive impairment in patients aged 50 and older. We developed prompts for two LLMs, Llama 2 and GPT-4, on Health Insurance Portability and Accountability Act (HIPAA)-compliant cloud-computing platforms using multiple approaches (e.g., hard prompting, retrieval augmented generation, and error analysis-based instructions) to select the optimal LLM-based method. Baseline models included a hierarchical attention-based neural network and XGBoost. Subsequently, we constructed an ensemble of the three models using a majority vote approach. Confusion-matrix-based scores were used for model evaluation. FindingsWe used a randomly annotated sample of 4,949 note sections from 1,969 patients (women: 1,046 [53.1%]; age: mean, 76.0 [SD, 13.3] years), filtered with keywords related to cognitive functions, for model development. For testing, a random annotated sample of 1,996 note sections from 1,161 patients (women: 619 [53.3%]; age: mean, 76.5 [SD, 10.2] years) without keyword filtering was utilised. GPT-4 demonstrated superior accuracy and efficiency compared to Llama 2, but did not outperform traditional models. The ensemble model outperformed the individual models in terms of all evaluation metrics with statistical significance (p<0.01), achieving a precision of 90.2% [95% CI: 81.9%-96.8%], a recall of 94.2% [95% CI: 87.9%-98.7%], and an F1-score of 92.1% [95% CI: 86.8%-96.4%]. Notably, the ensemble model showed a significant improvement in precision, increasing from a range of 70%-79% to above 90%, compared to the best-performing single model. Error analysis revealed that 63 samples were incorrectly predicted by at least one model; however, only 2 cases (3.2%) were mutual errors across all models, indicating diverse error profiles among them. InterpretationLLMs and traditional machine learning models trained using local EHR data exhibited diverse error profiles. The ensemble of these models was found to be complementary, enhancing diagnostic performance. Future research should investigate integrating LLMs with smaller, localised models and incorporating medical data and domain knowledge to enhance performance on specific tasks.

Reconstructing MODIS normalized difference snow index product on Greenland ice sheet using spatiotemporal extreme gradient boosting model

The spatiotemporally continuous data of normalized difference snow index (NDSI) are key to understanding the mechanisms of snow occurrence and development as well as the patterns of snow distribution changes. However, the presence of clouds, particularly prevalent in polar regions such as the Greenland Ice Sheet (GrIS), introduces a significant number of missing pixels in the MODIS NDSI daily data. To address this issue, this study proposes the utilization of a spatiotemporal extreme gradient boosting (STXGBoost) model generate a comprehensive NDSI dataset. In the proposed model, various input variables are carefully selected, encompassing terrain features, geometry-related parameters, and surface property variables. Moreover, the model incorporates spatiotemporal variation information, enhancing its capacity for reconstructing the NDSI dataset. Verification results demonstrate the efficacy of the STXGBoost model, with a coefficient of determination of 0.962, root mean square error of 0.030, mean absolute error of 0.011, and negligible bias (0.0001). Furthermore, simulation comparisons involving missing data and cross-validation with Landsat NDSI data illustrate the model’s capability to accurately reconstruct the spatial distribution of NDSI data. Notably, the proposed model surpasses the performance of traditional machine learning models, showcasing superior NDSI predictive capabilities. This study highlights the potential of leveraging auxiliary data to reconstruct NDSI in GrIS, with implications for broader applications in other regions. The findings offer valuable insights for the reconstruction of NDSI remote sensing data, contributing to the further understanding of spatiotemporal dynamics in snow-covered regions.