Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “data classification”

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 469 records · Page 26
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

Smart Methane Emission Detection System Development (Final Report)

Working with the Department of Energy's National Energy Technology Laboratory, Southwest Research Institute® (SwRI®) developed a system to identify methane leaks reliably, accurately, and autonomously at critical midstream sections of the natural gas distribution network in real-time for the purpose of mitigating methane emissions using Optical Gas Imaging (OGI) cameras. SwRI's Smart Leak Detection – Methane (SLED/M) adds a high degree of automation to the process of methane leak detection to minimize sources of human error, minimize response time to a leak event, and maximize midstream visibility. Furthermore, SwRI has been working towards integrating Quantitative OGI (QOGI) capabilities into this existing technology. By leveraging Deep Learning, SwRI now has the capability to estimate fugitive emission leak rates quickly and reliably, which allows operators to detect emissions, quantify leak rate, prioritize repairs, and validate the repairs in a single instrument. The next generation QOGI technology leverages the same cameras used in Leak Detection and Repair (LDAR) programs, with improvements in safety and speed for traditional quantification-based repairs, ultimately leading to less overhead cost for the operators. The goals for this research were to develop two types of models with the following goals: Run in real-time on the edge (≥ 12 Hz), Classification: Achieve less than 5% false positive detection, Classification: Achieve ≥ 95% methane plume detection rate, Regression: achieve ≤ 10 standard cubic feet per hour (scfh) prediction > 70% of the time. In order to achieve these results, multiple infrared (IR) and other sensors were investigated in tandem with the midwave IR (MWIR) OGI to provide additional information to train the underlying models. Information on atmospheric conditions including humidity, temperature, pressure, and solar radiation was provided by a weather station. Several machine learning and deep learning architectures and methods, including looking at quantized classification networks and regressions networks, were explored. As further data was collected, curated, and labeled, it allowed for more refined regressive networks to be adequately trained, leading to better insight into the true flow rates being observed. An important valuable deliverable of this research effort was the development of an advanced network which underwent multiple iterations capable of giving a continuous output. The current network has a predicted mean average percentage error (MAPE) of 12.3% just outside our target goal of 10.00%, but an accuracy of 97.78% at ±50 scfh, well within the overall goal for the Department of Energy (DOE) program. Upon closer inspection, it was observed that more than 10% of datapoints contributing to the MAPE predictions were the result of low flow rate predictions and are beyond the sensitivity of instrument measurement as a result of normal operational variation and noise.

03 NATURAL GAS↗
DOE OSTI2026

Post Wildfire Soil Bacterial MAGs and Metagenome Analysis

We compare the differences between bacteria in soil affected by a wildfire to an unaffected area from Minnewaska State Park, NY located in the biodiverse northern Shawangunk ridge. We detail our metagenomic sequencing data, relative abundance of bacterial phyla, and the taxonomic classification of three high-quality MAGs.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2021

Smart Methane Emission Detection System Development (Final Report)

Working with the Department of Energy’s National Energy Technology Laboratory, Southwest Research Institute® (SwRI®) developed a system to identify methane leaks reliably, accurately, and autonomously at critical midstream sections of the natural gas distribution network in real- time for the purpose of mitigating methane emissions using Optical Gas Imaging (OGI) cameras. SwRI’s Smart Leak Detection – Methane (SLED/M) adds a high degree of automation to the process of methane leak detection to minimize sources of human error, minimize response time to a leak event, and maximize midstream visibility. Furthermore, SwRI has been working towards integrating Quantitative OGI (QOGI) capabilities into this existing technology. By leveraging Deep Learning, SwRI now has the capability to estimate fugitive emission leak rates quickly and reliably, which allows operators to detect emissions, quantify leak rate, prioritize repairs, and validate the repairs in a single instrument. The next generation QOGI technology leverages the same cameras used in Leak Detection and Repair (LDAR) programs, with improvements in safety and speed for traditional quantification-based repairs, ultimately leading to less overhead cost for the operators. The goals for this research were to develop two types of models with the following goals: 1. Run in real-time on the edge (≥ 12 Hz) 2. Classification: Achieve less than 5% false positive detection 3. Classification: Achieve ≥ 95% methane plume detection rate 4. Regression: achieve ≤ 10 standard cubic feet per hour (scfh) prediction > 70% of the time In order to achieve these results, multiple infrared (IR) and other sensors were investigated in tandem with the midwave IR (MWIR) OGI to provide additional information to train the underlying models. Information on atmospheric conditions including humidity, temperature, pressure, and solar radiation was provided by a weather station. Several machine learning and deep learning architectures and methods, including looking at quantized classification networks and regressions networks, were explored. As further data was collected, curated, and labeled, it allowed for more refined regressive networks to be adequately trained, leading to better insight into the true flow rates being observed. An important valuable deliverable of this research effort was the development of an advanced network which underwent multiple iterations capable of giving a continuous output. The current network has a predicted mean average percentage error (MAPE) of 12.3% just outside our target goal of 10.00%, but an accuracy of 97.78% at ±50 scfh, well within the overall goal for the Department of Energy (DOE) program. Upon closer inspection, it was observed that more than 10% of datapoints contributing to the MAPE predictions were the result of low flow rate predictions and are beyond the sensitivity of instrument measurement as a result of normal operational variation and noise.

03 NATURAL GAS↗
DOE OSTI2020

High-resolution Hindcasts for U.S. Wave Energy Resource Characterization

The marine and hydrokinetic (MHK) industry is at an early stage of development and has the potential to play a significant role in diversifying the U.S. energy portfolio and reducing the U.S. carbon footprint. Wave energy is the largest among all the U.S. MHK energy resources, which include wave energy, ocean current, tidal-instream, ocean thermal energy conversion, and river-instream. Wave resource characterization is an essential step for regional wave energy assessments, Wave Energy Converter (WEC) project development, site selection and WEC design. The present paper provides an overview of a joint modelling effort by the Pacific Northwest National Laboratory and Sandia National Laboratories on high-resolution wave hindcasts to support the U.S. Department of Energy’s Water Power Technologies Office’s program of wave resource characterization, assessment and classifications in all US coastal regions. Topics covered include the modelling approach, model input requirements, model validation strategies, high performance computing resource requirements, model outputs and data management strategies. Examples of model setup and validation for different regions are provided along with application to development of classification systems, and analysis of regional wave climates. Lessons learned and technical challenges of the long-term, high-resolution regional wave hindcast are discussed.

Yang, Zhaoqing↗
DOE OSTI2020

Critical Infrastructure Classification via CNN-based Modeling and Image Analysis

With recent advances in the fields of satellite imagery and machine learning we now have the ability to develop explainable deep learning models that enhance critical infrastructure analysis. Funded through Idaho National Laboratory’s (INL) Laboratory Directed Research and Development (LDRD) office we are in the process of developing a deep learning model capable of identifying critical infrastructure facilities and embedded features within those facilities. Utilizing current limit of practice techniques in the machine learning areas of explainability and transfer learning our model, once complete, will have the capacity to be used on multiple different imagery data sets and produce results that not only classify critical infrastructure facilities, but also explain why a critical infrastructure facility was classified as a certain type of facility. These advancements eliminate the ‘black box’ approach deep learning models have had in the past, where a user will have to trust the conclusion of a model without understanding what reasoning when into the model’s classification process. They also expand a model’s usefulness, traditionally a deep learning model will have to use the same data set it was originally trained on. Given the long training times of deep learning models this is impractical in a number of scenarios. By utilizing transfer learning advancements, we are eliminating the need to train our model on the same data set it is then run on to classify critical infrastructure facilities. We are also enabling the analysis and classification of data sets that are potentially too small to be divided into a training and testing data set. Once completed this model can provide a foundation to enhanced critical infrastructure analysis, dependency analysis, and potential disaster relief efforts.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Effects of forest degradation classification on the uncertainty of aboveground carbon estimates in the Amazon

Tropical forests are critical for the global carbon budget, yet they have been threatened by deforestation and forest degradation by fire, selective logging, and fragmentation. Existing uncertainties on land cover classification and in biomass estimates hinder accurate attribution of carbon emissions to specific forest classes. In this study, we used textural metrics derived from PlanetScope images to implement a probabilistic classification framework to identify intact, logged and burned forests in three Amazonian sites. We also estimated biomass for these forest classes using airborne lidar and compared biomass uncertainties using the lidar-derived estimates only to biomass uncertainties considering the forest degradation classification as well. Our classification approach reached overall accuracy of 0.86, with accuracy at individual sites varying from 0.69 to 0.93. Logged forests showed variable biomass changes, while burned forests showed an average carbon loss of 35%. We found that including uncertainty in forest degradation classification significantly increased uncertainty and decreased estimates of mean carbon density in two of the three test sites. Our findings indicate that the attribution of biomass changes to forest degradation classes needs to account for the uncertainty in forest degradation classification. By combining very high-resolution images with lidar data, we could attribute carbon stock changes to specific pathways of forest degradation. This approach also allows quantifying uncertainties of carbon emissions associated with forest degradation through logging and fire. Both the attribution and uncertainty quantification provide critical information for national greenhouse gas inventories.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2021

Outlier classification using autoencoders: application for fluctuation driven flows in fusion plasmas

Understanding the statistics of fluctuation driven flows in the boundary layer of magnetically confined plasmas is desired to accurately model the lifetime of the vacuum vessel components. Mirror Langmuir probes (MLPs) are a novel diagnostic that uniquely allow us to sample the plasma parameters on a time scale shorter than the characteristic time scale of their fluctuations. Sudden large-amplitude fluctuations in the plasma degrade the precision and accuracy of the plasma parameters reported by MLPs for cases in which the probe bias range is of insufficient amplitude. While some data samples can readily be classified as valid and invalid, we find that such a classification may be ambiguous for up to 40% of data sampled for the plasma parameters and bias voltages considered in this study. In this contribution, we employ an autoencoder (AE) to learn a low-dimensional representation of valid data samples. By definition, the coordinates in this space are the features that mostly characterize valid data. Ambiguous data samples are classified in this space using standard classifiers for vectorial data. In this way, we avoid defining complicated threshold rules to identify outliers, which require strong assumptions and introduce biases in the analysis. By removing the outliers that are identified in the latent low-dimensional space of the AE, we find that the average conductive and convective radial heat fluxes are between approximately 5% and 15% lower as when removing outliers identified by threshold values. For contributions to the radial heat flux due to triple correlations, the difference is up to 40%.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2023

Exploring Classification of Topological Priors With Machine Learning for Feature Extraction

In many scientific endeavors, increasingly abstract representations of data allow for new interpretive methodologies and conceptualization of phenomena. For example, moving from raw imaged pixels to segmented and reconstructed objects allows researchers new insights and means to direct their studies toward relevant areas. Thus, the development of new and improved methods for segmentation remains an active area of research. With advances in machine learning and neural networks, scientists have been focused on employing deep neural networks such as U-Net to obtain pixel-level segmentations, namely, defining associations between pixels and corresponding/referent objects and gathering those objects afterward. Topological analysis, such as the use of the Morse-Smale complex to encode regions of uniform gradient flow behavior, offers an alternative approach: first, create geometric priors, and then apply machine learning to classify. This approach is empirically motivated since phenomena of interest often appear as subsets of topological priors in many applications. Using topological elements not only reduces the learning space but also introduces the ability to use learnable geometries and connectivity to aid the classification of the segmentation target. Here, in this article, we describe an approach to creating learnable topological elements, explore the application of ML techniques to classification tasks in a number of areas, and demonstrate this approach as a viable alternative to pixel-level classification, with similar accuracy, improved execution time, and requiring marginal training data.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Characterization and differentiation of aluminum powders used in improvised explosive devices. Part 2: Micromorphometric method refinement and preliminary statistical analysis

Abstract Aluminum (Al) powder is commonly encountered in improvised explosive devices (IEDs) as a metallic fuel due to its availability and low cost. Although available commercially in powder form, amateur bomb‐makers also produce their own Al powder via simple methods found online. In order to provide investigative leads and forensic intelligence, it is important to evaluate not only the composition of homemade devices, but also to distinguish between the various forms of Al powder they contain. To achieve this goal, a method using automated microscopy in combination with statistical techniques has been demonstrated to have the potential to provide source discrimination and investigative leads in source attribution of Al powders in IEDs. The present research refined this method and investigated 59 industrially and amateurly produced Al powder sources with seven subsamples per source using two traditional linear discriminant analyses (LDA), one with a standard data split for training and testing, and another using leave‐one‐out cross‐validation. Averaging the classification accuracies for the two LDA‐based analyses, LDA has the ability to correctly classify 59.26%, 83.35%, and 80.69% of the samples based on their powder source, type, and production method, respectively. This classification accuracy represents a 3407%, 317%, and 61.38% increase in accuracy from random class assignment, respectively. Further, in most instances of incorrect data attribution to a particular source, the subsample has been misidentified with another sample of the same powder type or production method.

Ommen, Danica M.↗
DOE OSTI2023

Galaxy Cruise: Deep Insights into Interacting Galaxies in the Local Universe

Abstract We present the first results from GALAXY CRUISE, a community (or citizen) science project based on data from the Hyper Suprime-Cam Subaru Strategic Program (HSC-SSP). The current paradigm of galaxy evolution suggests that galaxies grow hierarchically via mergers, but our observational understanding of the role of mergers is still limited. The data from HSC-SSP are ideally suited to improve our understanding with improved identifications of interacting galaxies thanks to the superb depth and image quality of HSC-SSP. We launched a community science project, GALAXY CRUISE, in 2019 and have collected over two million independent classifications of 20686 galaxies at z < 0.2. We first characterize the accuracy of the participants’ classifications and demonstrate that it surpasses previous studies based on shallower imaging data. We then investigate various aspects of interacting galaxies in detail. We show that there is a clear sign of enhanced activities of super-massive black holes and star formation in interacting galaxies compared to those in isolated galaxies. The enhancement seems particularly strong for galaxies undergoing violent mergers. We also show that the mass growth rate inferred from our results is roughly consistent with the observed evolution of the stellar mass function. The second season of GALAXY CRUISE is currently underway and we conclude with future prospects. We make the morphological classification catalog used in this paper publicly available at the GALAXY CRUISE website, which will be particularly useful for machine-learning applications.

Tanaka, Masayuki↗
DOE OSTI2022

Classification of Wetland Vegetation Based on NDVI Time Series from the HLS Dataset

Natural wetlands are intrinsically heterogeneous and typically composed of a mosaic of ecosystem patches with different vegetation types. Hydrological and biogeochemical processes in wetlands vary strongly among these ecosystem patches. To date, most remote sensing classification approaches for wetland vegetation either rely on coarse images that cannot capture the spatial variability of wetland vegetation or rely on very-high-resolution multi-spectral images that are detailed but very sporadic in time (less than once per year). This study aimed to use NDVI time series, generated from NASA’s HLS dataset, to classify vegetation patches. We demonstrate our approach at a temperate, coastal lake, estuarine marsh. To classify vegetation patches, a standard time series library of the four land-cover patch types was built from referencing specific locations that were identified as “pure” pixels. These were identified using a single-time high-resolution image. We calculated the distance between the HLS-NDVI time series at each pixel and the “pure”-pixel standards for each land-cover type. The resulting true-positive classified rate was >73% for all patch types other than water lily. The classification accuracy was higher in pixels of a more uniform composition. A set of vegetation maps was created for the years 2016 to 2020 at our research site to identify the vegetation changes at the site as it is affected by rapid water elevation increases in Lake Erie. Our results reveal how changes in water elevation have changed the patch distribution in significant ways, leading to the local extinction of cattail by 2019 and a continuous increase in the area cover of water lily patches.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2026

Dark Energy Survey Year 6 Results: Photometric Dataset for Cosmology

We describe the photometric dataset assembled from the full 6 yr of observations by the Dark Energy Survey (DES) in support of static-sky cosmology analyses. DES Y6 Gold is a curated dataset derived from DES Data Release 2 (DR2) that incorporates improved measurement, photometric calibration, object classification and value-added information. Y6 Gold comprises nearly 5000 deg$^{2}$ of grizY imaging in the south Galactic cap and includes 669 million objects with a depth of i$_{AB}$ ∼ 23.4 mag at a signal-to-noise ratio ∼ 10 for extended objects and a top-of-the-atmosphere photometric uniformity <2 mmag. Y6 Gold augments DES DR2 with simultaneous fits to multiepoch photometry for more robust galaxy shapes, colors, and photometric redshift estimates. Y6 Gold features improved morphological star–galaxy classification with an efficiency of 98.6% and a contamination of 0.8% for galaxies with 17.5 < i$_{AB}$ < 22.5. Additionally, it includes per-object quality information, and accompanying maps of the footprint coverage, masked regions, imaging depth, survey conditions, and astrophysical foregrounds that are used for cosmology analyses. After quality selections, benchmark samples contain 448 million galaxies and 120 million stars. This publication is complemented by data access and documentation.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2023

Dimensionality reduction using elastic measures

With the recent surge in big data analytics for hyperdimensional data, there is a renewed interest in dimensionality reduction techniques. In order for these methods to improve performance gains and understanding of the underlying data, a proper metric needs to be identified. This step is often overlooked, and metrics are typically chosen without consideration of the underlying geometry of the data. Here, in this paper, we present a method for incorporating elastic metrics into the t-distributed stochastic neighbour embedding (t-SNE) and Uniform Manifold Approximation and Projection (UMAP). We apply our method to functional data, which is uniquely characterized by rotations, parameterization and scale. If these properties are ignored, they can lead to incorrect analysis and poor classification performance. Through our method, we demonstrate improved performance on shape identification tasks for three benchmark data sets (MPEG-7, Car data set and Plane data set of Thankoor), where we achieve 0.77, 0.95 and 1.00 F1 score, respectively.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

MONTI: A Multi-Omics Non-negative Tensor Decomposition Framework for Gene-Level Integrative Analysis

Multi-omics data is frequently measured to enrich the comprehension of biological mechanisms underlying certain phenotypes. However, due to the complex relations and high dimension of multi-omics data, it is difficult to associate omics features to certain biological traits of interest. For example, the clinically valuable breast cancer subtypes are well-defined at the molecular level, but are poorly classified using gene expression data. Here, we propose a multi-omics analysis method called MONTI (Multi-Omics Non-negative Tensor decomposition for Integrative analysis), which goal is to select multi-omics features that are able to represent trait specific characteristics. Here, we demonstrate the strength of multi-omics integrated analysis in terms of cancer subtyping. The multi-omics data are first integrated in a biologically meaningful manner to form a three dimensional tensor, which is then decomposed using a non-negative tensor decomposition method. From the result, MONTI selects highly informative subtype specific multi-omics features. MONTI was applied to three case studies of 597 breast cancer, 314 colon cancer, and 305 stomach cancer cohorts. For all the case studies, we found that the subtype classification accuracy significantly improved when utilizing all available multi-omics data. MONTI was able to detect subtype specific gene sets that showed to be strongly regulated by certain omics, from which correlation between omics types could be inferred. Furthermore, various clinical attributes of nine cancer types were analyzed using MONTI, which showed that some clinical attributes could be well explained using multi-omics data. We demonstrated that integrating multi-omics data in a gene centric manner improves detecting cancer subtype specific features and other clinical features, which may be used to further understand the molecular characteristics of interest. The software and data used in this study are available at: https://github.com/inukj/MONTI.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2022

Photometric Classification of Early-time Supernova Light Curves with SCONE

Abstract In this work, we present classification results on early supernova light curves from SCONE, a photometric classifier that uses convolutional neural networks to categorize supernovae (SNe) by type using light-curve data. SCONE is able to identify SN types from light curves at any stage, from the night of initial alert to the end of their lifetimes. Simulated LSST SNe light curves were truncated at 0, 5, 15, 25, and 50 days after the trigger date and used to train Gaussian processes in wavelength and time space to produce wavelength–time heatmaps. SCONE uses these heatmaps to perform six-way classification between SN types Ia, II, Ibc, Ia-91bg, Iax, and SLSN-I. SCONE is able to perform classification with or without redshift, but we show that incorporating redshift information improves performance at each epoch. SCONE achieved 75% overall accuracy at the date of trigger (60% without redshift), and 89% accuracy 50 days after trigger (82% without redshift). SCONE was also tested on bright subsets of SNe ( r < 20 mag) and produced 91% accuracy at the date of trigger (83% without redshift) and 95% five days after trigger (94.7% without redshift). SCONE is the first application of convolutional neural networks to the early-time photometric transient classification problem. All of the data processing and model code developed for this paper can be found in the SCONE software package 1 1 github.com/helenqu/scone located at github.com/helenqu/scone (Qu 2021).

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2021

Drop Size Distribution Variability in Central Argentina during RELAMPAGO-CACTI

The Remote sensing of Electrification, Lightning, And Meso-scale/micro-scale Processes with Adaptive Ground Observations (RELAMPAGO) and the Cloud, Aerosol, and Complex Terrain Interactions Experiment Proposal (CACTI) field campaigns provided an unprecedented thirteen-disdrometer dataset in Central Argentina during the Intensive (IOP, 15 November to 15 December 2018) and Extended (EOP, 15 October 2018 to 30 April 2019) Observational Periods. The drop size distribution (DSD) parameters and their variability were analyzed across the region of interest, which was divided into three subregions characterized by the differing proximity to the Sierras de Córdoba (SDC), in order to assess the impact of complex terrain on the DSD parameters. A rigorous quality control of the data was first performed. The frequency distributions of DSD-derived parameters were analyzed, including the normalized intercept parameter (logN w ), the mean volume diameter (D 0 ), the mean mass diameter (D m ), the shape parameter (μ), the liquid water content (LWC), and the rain rate (R). The region closest to the SDC presented higher values of logN w , lower D 0 , and higher μ, while the opposite occurred in the farthest region, i.e., the concentration of small drops decreased while the concentration of bigger drops increased with the distance to the east of the SDC. Furthermore, the region closest to the SDC showed a bimodal distribution of D 0 : the lower values of D 0 were associated with higher values of logN w and were found more frequently during the afternoon, while the higher D 0 were associated with lower logN w and occurred more frequently during the night. The data were analyzed in comparison to the statistical analysis of Dolan et al. 2018 and sorted according to the classification proposed in the cited study. The logN w -D 0 and LWC-D 0 two-dimensional distributions allowed further discussion around the applicability of other mid-latitude and global precipitation classification schemes (startiform/convection) in the region of interest. Finally, three precipitation case studies were analyzed with supporting polarimetric radar data in order to relate the DSD characteristics to the precipitation type and the microphysical processes involved in each case.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2024

RECON Label Quality Report

The final quality of any AI/ML system is directly related to the quality of the input data used to train the system. In this case, we are trying to build a reliable image classifier that can correctly identify electrical components in x-ray images. The classification confidence is directly related to the quality of the labels in the training data, which are used in developing the AI/ML classifier. Incorrect or incomplete labels can substantially hinder the performance of the system during the training process, as it tries to compensate for variations that should not exist. Image labels are entered by subject matter experts, and in general can be assumed to be correct. However, this is not a guarantee, so developing ways to measure label quality and help identify or reject bad labels is important, especially as the database continues to grow. Given the current size of the database, a full manual review of each component is not feasible. This report will highlight the current state of the “RECON” x-ray image database and summarize several recent developments to try to help ensure high quality labeling both now and in the future. Questions that we hope to answer with this development include: 1) Are there any components with incorrect labels? 2) Can we suggest labels for components that are marked “Unknown”? 3) What kind of overall confidence do we have in the quality of the existing labels? 4) What systems or procedures can we put in place to maximize label quality?

42 ENGINEERING↗
DOE OSTI2026

Leveraging large language models to address data scarcity in machine learning for graphene synthesis

Machine learning in experimental materials science faces significant challenges due to the scarcity of data, which are costly and time-consuming to generate, particularly when relying on in-house experiments. Literature data mining offers a potential solution but introduces issues like mixed data quality, inconsistent formats, and non-uniform reporting of synthesis parameters, resulting in partially missing and heterogeneous features across the dataset. Here, we propose data imputation and feature engineering methods that employ pre-trained large language models (LLMs) to enhance machine learning performance on scarce, heterogeneous datasets, demonstrated on graphene CVD synthesis data and the ML-HydPARK hydrogen storage dataset. GPT models perform data imputation via tailored prompting and semantic normalization of inconsistently reported features through embeddings, for example, to harmonize the complex nomenclature of CVD substrates. Beyond yielding more diverse and richer feature representations than traditional methods such as K-nearest neighbors (KNN) and Multivariate Imputation by Chained Equations (MICE), LLM-based data imputation is evaluated against dataset characteristics and prompting strategies. We vary the level of autonomy granted to the LLM, from generic prompting that leverages pre-trained knowledge for autonomous data generation to data-informed prompting that constrains outputs using target-specific information, and demonstrate which level of autonomy yields superior imputation performance across datasets and feature types. The proposed data engineering methods markedly improve downstream performance; for example, in graphene layer number classification using a support vector machine (SVM), binary accuracy increases from 39% to 65% and ternary accuracy from 52% to 72%. Fine-tuning experiments on both datasets show that combining our proposed LLM-based data imputation and feature encoding methods with numerical machine learning predictors outperforms standalone fine-tuned LLM predictors in data-scarce settings. The proposed strategies emphasize data enhancement techniques rather than refining learning architectures or regularizing loss functions, offering a broadly applicable framework for improving machine learning performance on scarce, inhomogeneous datasets.

Chemical vapor deposition↗