Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Random forests”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 289 records · Page 16
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

Gap-filling eddy covariance methane fluxes: Comparison of machine learning model predictions and uncertainties at FLUXNET-CH4 wetlands

Time series of methane fluxes measured by eddy-covariance require gap-filling to estimate annual emissions. Gap-filling methane fluxes is challenging because of high variability and complex responses to multiple drivers. To date, there is no widely established gap-filling standard for methane, with regards both to the best model algorithms and predictors. In this study, we address the need for standardization by synthesizing results of gap-filling methods applied at 17 wetland sites spanning boreal to tropical regions including all major wetlands classes and two rice paddies. We introduce new procedures for: 1) creating realistic artificial gap scenarios, 2) training and evaluating gap-filling models without overstating performance, and 3) predicting half-hourly methane fluxes and annual emissions with robust uncertainty estimates. We tested a conventional method (marginal distribution sampling) and four machine learning algorithms - penalized linear regression, artificial neural networks, random forests, and boosted decision trees - and four predictor sets, including temporal, meteorological, ecosystem carbon and energy flux, and soil predictors. We find that the conventional method can achieve similar median performance to the machine learning models but is worse than the best machine learning models and relatively insensitive to predictor choices. Of the machine learning models, decision tree algorithms performed the best in cross-validation experiments, even with a baseline predictor set, and artificial neural networks showed comparable performance when using all predictors. Soil temperature was frequently the most important predictor whilst water table depth was important at sites with substantial water table fluctuations, highlighting the value of data on soil conditions. Raw gap-filling uncertainties from the machine learning models were underestimated and we propose a method to calibrate uncertainties to observations. Finally, we gap-fill and provide summary evaluation metrics for all 81 sites in the FLUXNET-CH4 community dataset and publicly release the python code for model development, evaluation, and uncertainty estimation.

42 ENGINEERING↗
DOE OSTI2020

A novel improved model for building energy consumption prediction based on model integration

Building energy consumption prediction plays an irreplaceable role in energy planning, management, and conservation. Constantly improving the performance of prediction models is the key to ensuring the efficient operation of energy systems. Moreover, accuracy is no longer the only factor in revealing model performance, it is more important to evaluate the model from multiple perspectives, considering the characteristics of engineering applications. Based on the idea of model integration, this paper proposes a novel improved integration model (stacking model) that can be used to forecast building energy consumption. The stacking model combines advantages of various base prediction algorithms and forms them into “meta-features” to ensure that the final model can observe datasets from different spatial and structural angles. Two cases are used to demonstrate practical engineering applications of the stacking model. A comparative analysis is performed to evaluate the prediction performance of the stacking model in contrast with existing well-known prediction models including Random Forest, Gradient Boosted Decision Tree, Extreme Gradient Boosting, Support Vector Machine, and K-Nearest Neighbor. The results indicate that the stacking method achieves better performance than other models, regarding accuracy (improvement of 9.5%–31.6% for Case A and 16.2%–49.4% for Case B), generalization (improvement of 6.7%–29.5% for Case A and 7.1%-34.6% for Case B), and robustness (improvement of 1.5%–34.1% for Case A and 1.8%–19.3% for Case B). The proposed model enriches the diversity of algorithm libraries of empirical models.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2023

Evaluating the impact of wildfire smoke on solar photovoltaic production

There are growing needs to understand how extreme weather events impact the electrical grid. Renewable energy sources such as solar photovoltaics are expanding in use to help sustainably meet electricity demands. Wildfires and, notably, the widespread smoke resulting from them, are one such extreme event that can impair the performance of solar photovoltaics. However, isolating the impact that smoke has on photovoltaic energy production, separate from ambient conditions, can be difficult. In this work, we seek to understand and quantify the impacts of wildfire smoke on solar photovoltaic production within the Western United States. Our analysis focuses on the construction of a random forest regression model to predict overall solar photovoltaic production. The model is used to separate and quantify the impacts of wildfire smoke in particular. To do so, we fuse historical weather, solar photovoltaic energy production, and PM2.5 particulate matter (primary smoke pollutant) data to train and test our model. The additional weather data allows us to capture interactions between wildfire smoke and other ambient conditions, as well as to create a more powerful predictive model capable of better quantifying the impacts of wildfire smoke on its own. We find that solar PV energy production decreases 8.3% on average during high smoke days at PV sites as compared to similar conditions without smoke present. Finally, this work allows us to improve our understanding of the potential impact on photovoltaic-based energy production estimates due to wildfire events and can help inform grid and operational planning as solar photovoltaic penetration levels continue to grow.

14 SOLAR ENERGY↗
DOE OSTI2024

Analyzing the impact of design factors on solar module thermomechanical durability using interpretable machine learning techniques

Solar modules in utility-scale systems are expected to maintain decades of lifetime to rival conventional energy sources. However, cyclic thermomechanical loading often degrades their long-term performance, highlighting the importance of effective design to mitigate thermal expansion mismatches between module materials. Given the complex composition of solar modules, isolating the impact of individual components on overall durability remains a challenging task. In this work, we analyze a comprehensive data set that comprises bill-of-materials (BOM) and thermal cycling power loss from 251 distinct module designs to identify the predominant design factors and their impacts on the thermomechanical durability of modules. The methodology of our analysis combines machine learning modeling (random forest) and Shapley additive explanation (SHAP) to correlate design factors with power loss and interpret the model’s decision-making. The interpretation reveals that silicon type (monocrystalline or polycrystalline), encapsulant thickness, busbar numbers, and wafer thickness predominantly influence the degradation. With lower power loss of around 0.6% on average in the SHAP analysis, monocrystalline cells present better durability than polycrystalline cells. This finding is further substantiated by statistical testing on our raw data set. The SHAP analysis also demonstrates that while thicker encapsulants lead to reduced power loss, further increasing their thickness over around 0.6 to 0.7 mm does not yield additional benefits, particularly for the front side one. In addition, other important BOM features such as the number of busbars are analyzed. This study provides a blueprint for utilizing explainable machine learning techniques in a complex material system and can potentially guide future research on optimizing the design of solar modules.

14 SOLAR ENERGY↗
DOE OSTI2020

DeepMerge: Classifying high-redshift merging galaxies with deep neural networks

In this work, we investigate and demonstrate the use of convolutional neural networks (CNNs) for the task of distinguishing between merging and non-merging galaxies in simulated images, and for the first time at high redshifts (i.e. $z=2$). We extract images of merging and non-merging galaxies from the Illustris-1 cosmological simulation and apply observational and experimental noise that mimics that from the Hubble Space Telescope; the data without noise form a "pristine" data set and that with noise form a "noisy" data set. The test set classification accuracy of the CNN is $79\%$ for pristine and $76\%$ for noisy. The CNN outperforms a Random Forest classifier, which was shown to be superior to conventional one- or two-dimensional statistical methods (Concentration, Asymmetry, the Gini, $M_{20}$ statistics etc.), which are commonly used when classifying merging galaxies. We also investigate the selection effects of the classifier with respect to merger state and star formation rate, finding no bias. Finally, we extract Grad-CAMs (Gradient-weighted Class Activation Mapping) from the results to further assess and interrogate the fidelity of the classification model.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2024

Constraining Galaxy-Halo connection using machine learning

We investigate the potential of machine learning (ML) methods to model small-scale galaxy clustering for constraining Halo Occupation Distribution (HOD) parameters. Our analysis reveals that while many ML algorithms report good statistical fits, they often yield likelihood contours that are significantly biased in both mean values and variances relative to the true model parameters. This highlights the importance of careful data processing and algorithm selection in ML applications for galaxy clustering, as even seemingly robust methods can lead to biased results if not applied correctly. ML tools offer a promising approach to exploring the HOD parameter space with significantly reduced computational costs compared to traditional brute-force methods if their robustness is established. Using our ANN-based pipeline, we successfully recreate some standard results from recent literature. Properly restricting the HOD parameter space, transforming the training data, and carefully selecting ML algorithms are essential for achieving unbiased and robust predictions. Among the methods tested, artificial neural networks (ANNs) outperform random forests (RF) and ridge regression in predicting clustering statistics, when the HOD prior space is appropriately restricted. We demonstrate these findings using the projected two-point correlation function (w p (r p )), angular multipoles of the correlation function (ξ ℓ (r)), and the void probability function (VPF) of Luminous Red Galaxies from Dark Energy Spectroscopic Instrument mocks. Our results show that while combining w p (r p ) and VPF improves parameter constraints, adding the multipoles ξ 0 , ξ 2 , and ξ 4 to w p (r p ) does not significantly improve the constraints.

cosmology↗
DOE OSTI2022

Microglia are implicated in the development of paclitaxel chemotherapy-associated cognitive impairment in female mice

Chemotherapy remains a mainstay in the treatment of many types of cancer even though it is associated with debilitating behavioral side effects referred to as “chemobrain,” including difficulty concentrating and memory impairment. The predominant hypothesis in the field is that systemic inflammation drives these cognitive impairments, although the brain mechanisms by which this occurs remain poorly understood. Here, we hypothesized that microglia are activated by chemotherapy and drive chemotherapy-associated cognitive impairments. To test this hypothesis, we treated female C57BL/6 mice with a clinically-relevant regimen of a common chemotherapeutic, paclitaxel (6 i.p. doses at 30 mg/kg), which impairs memory of an aversive stimulus as assessed via a contextual fear conditioning (CFC) paradigm. In this work, paclitaxel increased the percent area of IBA1 staining in the dentate gyrus of the hippocampus. Moreover, using a machine learning random forest classifier we identified immunohistochemical features of reactive microglia in multiple hippocampal subregions that were distinct between vehicle- and paclitaxel-treated mice. Paclitaxel treatment also increased gene expression of inflammatory cytokines in a microglia-enriched population of cells from mice. Lastly, a selective inhibitor of colony stimulating factor 1 receptor, PLX5622, was employed to deplete microglia and then assess CFC performance following paclitaxel treatment. PLX5622 significantly reduced hippocampal gene expression of paclitaxel-induced proinflammatory cytokines and restored memory, suggesting that microglia play a critical role in the development of chemotherapy-associated neuroinflammation and cognitive impairments. This work provides critical evidence that microglia drive paclitaxel-associated cognitive impairments, a key mechanistic detail for determining preventative and intervention strategies for these burdensome side effects.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2021

Growing grasses in unprofitable areas of US Midwest croplands could increase species richness

The US has large potential to grow perennial energy crops, but because these crops are rarely grown in current agricultural landscapes, it is unclear how biodiversity may be affected. Over time, as agriculture has increased, many grassland species have declined. In addition, not all agricultural land is profitable for growing annual crops. Unprofitable areas were responsible for a loss of approximately $110 million USD per year from 2013 to 2016. Based on this, we want to know how converting less-profitable portions of agricultural fields to switchgrass, a native prairie grass, would influence species occurrence. To address this question, we developed an alternative landscape in which clustered corn/soy acres with a low return on investment (ROI) were replaced with grassland. We also developed and validated species distribution models to predict changes in species occurrence for 28 avian species in Iowa in response to landscape management. Furthermore, we compared results for three different models: Random forest (RF), Stochastic gradient boosting (GBM), and Neural network (Nnet) and found that all models performed well and predicted similar species distribution. Predicted species richness increased by 3.66% (RF), 2.79% (GBM), and 7.51% (Nnet) when we simulated a change in management for ~3% of Iowa's low ROI corn/soybean areas to grassland. If harvested, these areas could generate approximately 7.6 million dry tons/year of switchgrass for bioenergy, thereby increasing farmers earnings. Unprofitable areas tended to occur along streams, which suggest that incorporating partially harvested riparian buffers can benefit avian biodiversity, while improving water quality and reducing unnecessary costs for farmers.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2021

Learning-based CO 2 concentration prediction: Application to indoor air quality control using demand-controlled ventilation

There have been increasing concerns over the air quality inside buildings as high levels of bio-effluents can cause nausea, dizziness, headaches, and fatigue to the people working in those spaces. First published in 2004 as Standard 62.1, ASHRAE Standard 62.2-2019 requires highly occupied spaces to implement heating, ventilation, and air conditioning (HVAC) that can dilute contaminants produced by occupants. In this regard, occupant-centric ventilation control has been regarded as an effective practice to maintain a satisfactory indoor air quality (IAQ) when dealing with highly variable occupancy environments. However, few established models in current literature and practice consider dynamic occupancy behavior and adaptive IAQ control. To address this gap, a dynamic indoor CO2 model is constructed using machine learning algorithms to forecast CO2concentrations across a range of forecasting horizons. Herein, we tuned and compared six state-of-the-algorithms—including Support Vector Machine, Ada Boost, Random Forest, Gradient Boosting, Logistic Regression, and Multilayer Perceptron. The algorithms’ performances are validated using CO 2 and historical meteorological data collected from a campus classroom with a variable occupancy rate. Simulation results showed that Multilayer Perceptron can strongly predict the volatile CO 2 behavior and also outperforms other algorithms in terms of accuracy. Furthermore, a control strategy capable of modeling and detecting dynamic patterns of CO 2 level is utilized to modulate the ventilation rate in real-time and also reduce the energy consumption. The proposed controller reduced the HVAC fan’s energy consumption by 51.4% and provide ventilation as needed per the ASHRAE standards.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2022

Smart detection of indoor occupant thermal state via infrared thermography, computer vision, and machine learning

The ability to measure occupants’ thermal state in real time will enable major advances in the control of air conditioning systems. This study proposes predicting occupant thermal state by a combination of infrared thermography, computer vision, and machine learning. The approach (1) uses cheek, nose, and hand temperatures because they are least subject to blockage by hair, glasses, and clothing; (2) measures the distribution of skin temperatures within geometrically defined sub-areas of the face and hand; and (3) uses temperature differences within and between these areas to eliminate the effects of calibration drift that are unavoidable in thermal infrared (TIR) cameras. Two series of tests were conducted, respectively in an outdoor carport and an indoor environmental chamber, collecting a total of 48,422 sets of cheek, nose, and hand skin temperatures using a TIR camera and computer-vision technology, coupled with 715 subjective responses of thermal sensations. To predict occupant thermal state, Random Forest classification models were built using either absolute skin temperatures (the maximum and median temperatures of cheek and hand segments, and the temperature of the central spot on the nose), or intra- and inter-segment temperature differences of cheeks, hands, and nose. These measurements were found to accurately predict occupant thermal state. Using the maximum and median temperatures for cheek and nose, or for cheek and hand, predicts thermal state with an accuracy of 92–96%. In conclusion, using only the intra- and inter-segment temperature differences from cheek and nose is 83% accurate; adding the hand temperature differences increases the accuracy to 96%.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2021

Machine learning feature analysis illuminates disparity between E3SM climate models and observed climate change

In September of 2020, Arctic sea ice extent was the second-lowest on record. State of the art climate prediction uses Earth system models (ESMs), driven by systems of differential equations representing the laws of physics. Previously, these models have tended to underestimate Arctic sea ice loss. The issue is grave because accurate modeling is critical for economic, ecological, and geopolitical planning. We use machine learning techniques, including random forest regression and Gini importance, to show that the Energy Exascale Earth System Model (E3SM) relies too heavily on just one of the ten chosen climatological quantities to predict September sea ice averages. Furthermore, E3SM gives too much importance to six of those quantities when compared to observed data. Finally, identifying the features that climate models incorrectly rely on should allow climatologists to improve prediction accuracy.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2026

Cross-domain digital twin architecture for predictive maintenance via machine learning and Large Language Models

This research introduces a comprehensive framework for creating and deploying a digital twin platform for continuous monitoring and predictive maintenance within industrial settings. Through utilizing advanced technologies, including Unreal Engine 5, Unity 3D, the Message Queue Telemetry Transport protocol, Random Forest machine learning algorithms, and Large Language Models (LLMs), we establish a platform that digitally reproduces physical equipment and translates digital controls into real-world actions. This facilitates preventive maintenance approaches and improves operational effectiveness. The digital twin platform gathers sensor data from operational equipment, analyzes it using machine learning, and delivers practical insights to prevent potential malfunctions and enhance equipment performance. Furthermore, the incorporation of a web portal enables efficient monitoring and access to historical data, educational materials, and equipment status information. Preliminary findings indicate that digital twins can transform industrial equipment management and maintenance methodologies.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Machine learning elastic constants of multi-component alloys

The present manuscript explores application of machine learning methods for determining elastic constants and other derived mechanical properties of multi-component alloys. Here, a number of machine learning models, including linear regression, neural network and random forest based models, are trained and tested on a dataset of binary alloys generated using density functional theory (DFT) calculations and spanning over a large number of elemental species in the periodic table. Starting with a wide range of simple and easily accessible compositionally-averaged elemental features, a correlation-based feature selection strategy was used to systematically down-select a set of most relevant features towards the prediction of the elasticity tensor components. The true predictive performance and the associated uncertainties of the models were established by testing on unseen data and bootstrapping, respectively. A single and pair-wise feature partial dependence analysis was performed to visualize the average property trends in the multi-dimensional feature space in order to further understand the achieved predictive performance. The utility of the trained model is further demonstrated by obtaining sufficiently accurate yet highly efficient approximations for bulk modulus, Young’s modulus, shear modulus and Poisson’s ratio for alloys beyond the binary space (i.e., two-component alloys) on which the model was originally trained. More importantly, we test and validate the predictive performance of the developed model directly against the experimentally measured elastic constants of technologically relevant multi-component alloys (such as, Ni- and Ti-based alloys). Finally, utility of such a data-enabled route is demonstrated by predicting the possible range of various elastic properties for vast composition space available within the five component Ni-Cr-Fe-Mo-W alloy system in a high-throughput manner.

36 MATERIALS SCIENCE↗
DOE OSTI2023

Understanding and control of Zener pinning via phase field and ensemble learning

Zener pinning refers to the dispersion of fine particles which influences grain size distribution via movement of grain boundaries in a polycrystalline material. Grain size distribution in polycrystals has a significant impact on their properties including physical, chemical, mechanical, and optical to name a few. We explore the use of Phase-field modeling and machine-learning techniques to understand and improve the control of grain size distribution via Zener pinning in polycrystalline materials. We develop a machine learning model that determines the relative importance of various parameters to exercise microstructure control via Zener pinning. Our workflow combines high-throughput phase-field simulations and machine learning to address the computational bottlenecks associated with large-scale simulations as well as identify features necessary for microstructure control in polycrystals. A random forest (RF) regression model was developed to predict grain sizes based on five Phase-field model parameters, achieving an average prediction error of 0.72 nm for the training data and 1.44 nm for the test data. The importance of the input parameters is analyzed using the SHapley Additive exPlanations (SHAP) approach which reveals that diffusivity, volume fraction, and particle diameter are the most important parameters in determining the final grain size. These findings will allow us to select the best second-phase particles, optimize grain size distributions and thus design microstructures with the desired properties. The developed method is a highly versatile and generalizable approach that can be used to assess the combined effects of individual features in the presence of multiple variables.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Estimating soybean yields from high-temporal-resolution multi-source data using deep learning

Accurate and timely crop yield prediction is crucial for ensuring food security and maintaining stable agricultural markets. In recent years, there has been a surge in interest in leveraging high-temporal-resolution, multi-source data for effective crop growth monitoring and yield estimation. A notable challenge arises from the difficulty in capturing the intricate interactions between variables across different time steps within these high-temporal-resolution time series datasets. This complexity hinders the reliable extraction of yield information from voluminous and often noisy datasets, especially during periods of extreme weather events. Here, in this study, we propose an Attention and Graph Isomorphism Network-enhanced Bi-directional Long Short-Term Memory network (AGB-LSTM) for estimating county-level soybean yield in the United States. This model integrates a diverse set of remote sensing data, including Near-Infrared Reflectance of Vegetation (NIRv), Sun-Induced chlorophyll Fluorescence (SIF), and Gross Primary Productivity (GPP), along with environmental covariates. The AGB-LSTM effectively leverages information related to crop yield from high-temporal-resolution time series data (5-days), achieving an accuracy of R²= 0.67 and rRMSE = 14.46%. This approach significantly outperforms traditional machine learning methods such as Random Forest (RF) (R²= 0.52, rRMSE = 17.36%) and Bi-LSTM (R²= 0.58, rRMSE = 16.17%). Sensitivity experiments with different time steps and ranges demonstrated that our model could accurately and stably predict yields 1 to 2 months before harvest. Moreover, data with a finer temporal resolution consistently improved prediction performance, resulting in an approximately 20% increase in and an approximately 20% decrease in rRMSE compared to using monthly composites. We also evaluated the robustness of the model under extreme climate events and observed strong performance (R²= 0.50, rRMSE = 21.32%). Finally, yield mapping for major soybean-producing regions in North America in 2023 revealed spatial patterns that closely matched USDA yield reports. Our findings suggest that the AGB-LSTM model is a promising and effective method for estimating yield and has notable potential for global crop yield forecasting.

Deep learning↗
DOE OSTI2024

Graph neural networks for CO 2 solubility predictions in Deep Eutectic Solvents

Deep Eutectic Solvents (DESs) are a promising class of solvents for CO 2 capture. DESs are complex mixtures that can be designed to optimize CO solubility and overall capture process efficiency. However, the vast design landscape of DES mixtures makes experimental investigation prohibitive; as such, there is a need for computational models that can quickly and efficiently navigate the design space and inform data collection efforts. In this work, we propose Graph Neural Network (GNN) models for predicting CO 2 solubility for DESs; the GNN leverages a mixture graph representation that captures the molecular structure of the DES components as well as their intermolecular interactions. Here, we compare the GNN framework against alternative architectures (neural networks, graph convolution networks, and random forests) and data representations (molecular fingerprints, sigma profiles, and graphs). We show that the proposed approach offers superior predictive performance; specifically, we show that solubility can be predicted reliably directly from molecular structure (without the need of using sigma profiles as proposed in previous studies). This result is important, as obtaining sigma profiles requires expensive density functional theory computations. We also explored the ability of GNNs to predict solubility for new DES mixtures and operating conditions. We found that the model extrapolates across temperature reliably. However, we also found deficiencies in the ability of the model to predict solubility for DES mixtures, pressures, and molar ratio not included in the training sets; we show that this is due to an inherent lack of chemical diversity in datasets available in the literature. The proposed computational capabilities can thus help navigate the design space of DES and inform data collection efforts. Our models, data, and benchmarks are shared as Python code implemented in Jupyter notebooks.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2024

Machine learning in materials research: Developments over the last decade and challenges for the future

The number of studies that apply machine learning (ML) to materials science has been growing at a rate of approximately 1.67 times per year over the past decade. In this review, I examine this growth in various contexts. First, I present an analysis of the most commonly used tools (software, databases, materials science methods, and ML methods) used within papers that apply ML to materials science. The analysis demonstrates that despite the growth of deep learning techniques, the use of classical machine learning is still dominant as a whole. It also demonstrates how new research can effectively build upon past research, particular in the domain of ML models trained on density functional theory calculation data. Next, I present the progression of best scores as a function of time on the matbench materials science benchmark for formation enthalpy prediction. In particular, a dramatic improvement of 7 times reduction in error is obtained when progressing from feature-based methods that use conventional ML (random forest, support vector regression, etc.) to the use of graph neural network techniques. Finally, I provide views on future challenges and opportunities, focusing on data size and complexity, extrapolation, interpretation, access, and relevance.

36 MATERIALS SCIENCE↗
DOE OSTI2023

Predictive models of long COVID

Background: The cause and symptoms of long COVID are poorly understood. It is challenging to predict whether a given COVID-19 patient will develop long COVID in the future. Methods: We used electronic health record (EHR) data from the National COVID Cohort Collaborative to predict the incidence of long COVID. We trained two machine learning (ML) models — logistic regression (LR) and random forest (RF). Features used to train predictors included symptoms and drugs ordered during acute infection, measures of COVID-19 treatment, pre-COVID comorbidities, and demographic information. We assigned the ‘long COVID’ label to patients diagnosed with the U09.9 ICD10-CM code. The cohorts included patients with (a) EHRs reported from data partners using U09.9 ICD10-CM code and (b) at least one EHR in each feature category. We analysed three cohorts: all patients (n = 2,190,579; diagnosed with long COVID = 17,036), inpatients (149,319; 3,295), and outpatients (2,041,260; 13,741). Findings: LR and RF models yielded median AUROC of 0.76 and 0.75, respectively. Ablation study revealed that drugs had the highest influence on the prediction task. The SHAP method identified age, gender, cough, fatigue, albuterol, obesity, diabetes, and chronic lung disease as explanatory features. Models trained on data from one N3C partner and tested on data from the other partners had average AUROC of 0.75. Interpretation: ML-based classification using EHR information from the acute infection period is effective in predicting long COVID. SHAP methods identified important features for prediction. Cross-site analysis demonstrated the generalizability of the proposed methodology.

60 APPLIED LIFE SCIENCES↗