Search NASA⌕ Search

SEARCH · Search NASA

Results for “support vector regression”

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 181 records · Page 10
All sourcesNASA NTRSDOE OSTI
DOE OSTI2022

Sampling-based Sublinear Low-rank Matrix Arithmetic Framework for Dequantizing Quantum Machine Learning

We present an algorithmic framework for quantum-inspired classical algorithms on close-to-low-rank matrices, generalizing the series of results started by Tang’s breakthrough quantum-inspired algorithm for recommendation systems [STOC’19]. Motivated by quantum linear algebra algorithms and the quantum singular value transformation (SVT) framework of Gilyén et al. [STOC’19], we develop classical algorithms for SVT that run in time independent of input dimension, under suitable quantum-inspired sampling assumptions. Our results give compelling evidence that in the corresponding QRAM data structure input model, quantum SVT does not yield exponential quantum speedups. Since the quantum SVT framework generalizes essentially all known techniques for quantum linear algebra, our results, combined with sampling lemmas from previous work, suffice to generalize all prior results about dequantizing quantum machine learning algorithms. In particular, our classical SVT framework recovers and often improves the dequantization results on recommendation systems, principal component analysis, supervised clustering, support vector machines, low-rank regression, and semidefinite program solving. We also give additional dequantization results on low-rank Hamiltonian simulation and discriminant analysis. Our improvements come from identifying the key feature of the quantum-inspired input model that is at the core of all prior quantum-inspired results: ℓ 2 -norm sampling can approximate matrix products in time independent of their dimension. We reduce all our main results to this fact, making our exposition concise, self-contained, and intuitive.

Computer Science↗
DOE OSTI2021

Artificial Diversity and Defense Security (ADDSec)

Artificial Diversity and Defense Security (ADDSec) machine learning algorithms are used to classify and cluster threats so that an appropriate response can be initiated as a mitigation strategy. The package includes an ensemble of machine learning algorithms such as Support Vector Machines, naïve bayes, logistic regression, and random forest that evolve with the data to recognize anomalous behavior at the host and network levels. Inputs into the machine learning algorithms include end host system calls, system utilization, packet captures, and syslog messages. The machine learning algorithms can be retrained based on user defined intervals or on the number of packets received. ADDSEC's threat responses include Internet Protocol (IP) Address randomization, application port number randomization, and application library randomization. The IP randomization implementation is built on top of a Software Defined Networking (SDN) framework. The SDN controller installs flows on each of the SDN switches with randomized source and destination IP addresses. The application port numbers are randomized using iptables. The application library randomization is created with a LLVM compiler. All randomization schemes are transparent to the endpoints on the network. Sandia National Laboratories is a multimission laboratory managed and operated by National Technology & Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA0003525. SAND2021-3379 O

Cox, RebeccaE.↗
NASA NTRS2018

Predicting the Operational Acceptance of Airborne Flight Reroute Requests Using Data Mining

For tools that generate more efficient flight routes or reroute advisories, it is important to ensure compatibility of automation and autonomy decisions with human objectives so as to ensure acceptability by the human operators. In this paper, the authors developed a proof of concept predictor of operational acceptability for route changes during a flight. Such a capability could have applications in automation tools that identify more efficient routes around airspace impacted by weather or congestion and that better meet airline preferences. The predictor is based on applying data mining techniques, including logistic regression, a decision tree, a support vector machine, a random forest and Adaptive Boost, to historical flight plan amendment data reported during operations and field experiments. Cross validation was used for model development, while nested cross validation was used to validate the models. The model found to have the best performance in predicting air traffic controller acceptance or rejection of a route change, using the available data from Fort Worth Air Traffic Control Center and its adjacent Centers, was the random forest, with an F-score of 0.77. This result indicates that the operational acceptance of reroute requests does indeed have some level of predictability, and that, with suitable data, models can be trained to predict the operational acceptability of reroute requests. Such models may ultimately be used to inform route selection by decision support tools, contributing to the development of increasingly autonomous systems that are capable of routing aircraft with less human input than is currently the case.

Operational Acceptability↗
NASA NTRSDate not supplied

Classifying Agnostic Biosignatures using Raman, VNIR, and Elemental Data

How can we use our current wealth of terrestrial data, encompassing biogenic and abiogenic systems, to determine the distinguishing properties of life? SCOBI (Statistical Classification of Biosignature Information) uses machine learning techniques to algorithmically identify combinations of measurements that are “indicative of life”. A set of ~1000 observations, comprising elemental abundance, isotopic fractionation, VNIR reflectance, and (in progress) Raman spectra, have been assembled from existing literature and databases. The observations cover systems classified as “indicative alive” (e.g., cells, vegetation), “indicative non-alive” (e.g., fossils, teeth), “mixed indicative” (e.g., soil, pond water), or “non-indicative” (e.g., rocks, meteorites). VNIR data was preprocessed by linear interpolation from 400-2100 nm and smoothed with a Savitzky-Golay filter. To limit the amount of Earth-biochemistry-specific (non-agnostic) information included, the first five spectral features extracted were number of peaks, number of troughs, mean reflectance, mean peak width, and broadest peak width. To help further emphasize agnostic biosignatures, Earth-specific features such as chlorophylls have been manually flagged so that feature importance with and without them can be compared. Classifiers including k-nearest neighbors (KNN), Gaussian Naïve Bayes (GNB), logistic regression (LR), random forest (RF), and support vector machine (SVM) were implemented, as was a combination voting classifier. Performance metrics included false positive rates, false negative rates, and AUC with 50-50 test/train splits (Monte Carlo simulations). Key takeaways from this stage, prior to the inclusion of Raman spectra, are (1) the overall success rate of 0.933 AUC was most heavily influenced by the elemental abundance data; and (2) VNIR reflectance had the lowest classification performance with 0.52 AUC (58% of objects correctly classified). The next steps are to complete integration of Raman spectral data and to improve the approach to pre-processing and feature extraction for both types of spectral data, such as automated baseline removal, whole spectrum matching, and dimensionality reduction.

Biosignatures↗
DOE OSTI2024

Analysis of Waste Material Feedstocks Using Laser-Induced Breakdown Spectroscopy and Machine Learning

Predicting properties such as heating value, ash fusion temperature, and mineral ash composition from Laser-Induced Breakdown Spectroscopy (LIBS) data can make gasifiers more flexible to different feedstocks. Understanding these feedstock properties in-situ improves feedstock conversion modelling methods that allow for consistent operation, higher carbon conversion, and reduced fouling and erosion rates. The purpose of this study is to demonstrate methods for model creation that take LIBS data as predictor features and estimate higher order material properties as a function of feedstock material properties. Six samples were chosen to represent a mixture of abundant and carbon rich waste materials. LIBS measurements were performed on these samples for elemental wavelengths and intensity values. Laboratory analytical results were obtained for each sample’s heating value, proximate and ultimate analysis, mineral ash composition, ash fusion temperatures, and viscosity temperatures. Thermal conductivity was measured using a HotDisk TPS 2500S. LIBS measurements were processed and used as predictor features for machine learning (ML) models to predict the sample’s material properties. Predictor feature selection algorithms, particularly minimum redundancy maximum relevance (mRMR), reduced the dimensionality of ML models. Many modelling methods such as Gaussian process regression (GPR), regression tree, neural networks (NN), and support vector machines (SVM) were demonstrated to be effective at predicting higher order properties; however, mRMR with GPR stood out as a clear winning combination.

01 COAL, LIGNITE, AND PEAT↗
DOE OSTI2023

When less is more: How increasing the complexity of machine learning strategies for geothermal energy assessments may not lead toward better estimates

Previous moderate- and high-temperature geothermal resource assessments of the western United States utilized data-driven methods and expert decisions to estimate resource favorability. Although expert decisions can add confidence to the modeling process by ensuring reasonable models are employed, expert decisions also introduce human and, thereby, model bias. This bias can present a source of error that reduces the predictive performance of the models and confidence in the resulting resource estimates. Our study aims to develop robust data-driven methods with the goals of reducing bias and improving predictive ability. We present and compare nine favorability maps for geothermal resources in the western United States using data from the U.S. Geological Survey's 2008 geothermal resource assessment. Two favorability maps are created using the expert decision-dependent methods from the 2008 assessment (i.e., weight-of-evidence and logistic regression). With the same data, we then create six different favorability maps using logistic regression (without underlying expert decisions), XGBoost, and support-vector machines paired with two training strategies. The training strategies are customized to address the inherent challenges of applying machine learning to the geothermal training data, which have no negative examples and severe class imbalance. We also create another favorability map using an artificial neural network. We demonstrate that modern machine learning approaches can improve upon systems built with expert decisions. We also find that XGBoost, a non-linear algorithm, produces greater agreement with the 2008 results than linear logistic regression without expert decisions, because the expert decisions in the 2008 assessment rendered the otherwise linear approaches non-linear despite the fact that the 2008 assessment used only linear methods. The F1 scores for all approaches appear low (F1 score < 0.10), do not improve with increasing model complexity, and, therefore, indicate the fundamental limitations of the input features (i.e., training data). Until improved feature data are incorporated into the assessment process, simple non-linear algorithms (e.g., XGBoost) perform equally well or better than more complex methods (e.g., artificial neural networks) and remain easier to interpret.

15 GEOTHERMAL ENERGY↗
DOE OSTI2017

A Case Study on a Combination NDVI Forecasting Model Based on the Entropy Weight Method

It is critically meaningful to accurately predict NDVI (Normalized Difference Vegetation Index), which helps guide regional ecological remediation and environmental managements. In this study, a combination forecasting model (CFM) was proposed to improve the performance of NDVI predictions in the Yellow River Basin (YRB) based on three individual forecasting models, i.e., the Multiple Linear Regression (MLR), Artificial Neural Network (ANN), and Support Vector Machine (SVM) models. The entropy weight method was employed to determine the weight coefficient for each individual model depending on its predictive performance. Results showed that: (1) ANN exhibits the highest fitting capability among the four orecasting models in the calibration period, whilst its generalization ability becomes weak in the validation period; MLR has a poor performance in both calibration and validation periods; the predicted results of CFM in the calibration period have the highest stability; (2) CFM generally outperforms all individual models in the validation period, and can improve the reliability and stability of predicted results through combining the strengths while reducing the weaknesses of individual models; (3) the performances of all forecasting models are better in dense vegetation areas than in sparse vegetation areas.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2022

Machine learning models for rat multigeneration reproductive toxicity prediction

Reproductive toxicity is one of the prominent endpoints in the risk assessment of environmental and industrial chemicals. Due to the complexity of the reproductive system, traditional reproductive toxicity testing in animals, especially guideline multigeneration reproductive toxicity studies, take a long time and are expensive. Therefore, machine learning, as a promising alternative approach, should be considered when evaluating the reproductive toxicity of chemicals. We curated rat multigeneration reproductive toxicity testing data of 275 chemicals from ToxRefDB (Toxicity Reference Database) and developed predictive models using seven machine learning algorithms (decision tree, decision forest, random forest, k-nearest neighbors, support vector machine, linear discriminant analysis, and logistic regression). A consensus model was built based on the seven individual models. An external validation set was curated from the COSMOS database and the literature. The performances of individual and consensus models were evaluated using 500 iterations of 5-fold cross-validations and the external validation data set. The balanced accuracy of the models ranged from 58% to 65% in the 5-fold cross-validations and 45%–61% in the external validations. Prediction confidence analysis was conducted to provide additional information for more appropriate applications of the developed models. The impact of our findings is in increasing confidence in machine learning models. We demonstrate the importance of using consensus models for harnessing the benefits of multiple machine learning models (i.e., using redundant systems to check validity of outcomes). While we continue to build upon the models to better characterize weak toxicants, there is current utility in saving resources by being able to screen out strong reproductive toxicants before investing in vivo testing. The modeling approach (machine learning models) is offered for assessing the rat multigeneration reproductive toxicity of chemicals. Our results suggest that machine learning may be a promising alternative approach to evaluate the potential reproductive toxicity of chemicals.

consensus model↗
NASA NTRS2025

Improving Adhesive Bondline Time of Flight Predictions During Autoclave Cure Utilizing Machine Learning

Composite materials are increasingly being used in aerospace applications due to their superior strength-to-weight ratio compared to commonly used metals. A current limitation to widespread adoption is the certification of adhesively bonded joints. One approach to improving adhesive bonding in composites is accurately measuring the thickness of adhesive bondlines in composite laminates. Precise bondline thickness control is essential for aerospace applications where adhesive layer thickness directly affects joint fracture properties and structural performance. This study focused on implementing machine learning techniques to determine the ultrasonic time of flight (directly correlated to thickness) in adhesive bondlines throughout autoclave cure cycles. A high-temperature (use up to 180°C) ultrasonic scanning system was deployed in an autoclave to provide time of flight data through composite panels. Three experiments were conducted on the curing of 305 mm × 305 mm unidirectional composite panels. In the first experiment, a piecewise function was fit for the temperature correction factor to account for changing autoclave temperatures. Due to deficiencies in the first calibration experiment, a second experiment was run, and the results were used to train a machine learning model. The revised experiment, in combination with the machine learning model, significantly increased the accuracy of the bondline time of flight predictions (~14% error reduced to <1%). Data was processed using the Regression Learner Application in MATLAB®, with a Support Vector Machine selected for the model. The result was a machine learning algorithm capable of reliably quantifying ultrasonic time of flight through adhesive bondlines. The third experiment provided independent test data for the machine learning model, demonstrating that the model produces accurate predictions from data beyond its training set.

Machine Learning↗
DOE OSTI2021

Fault Detection Utilizing Convolution Neural Network on Timeseries Synchrophasor Data From Phasor Measurement Units

An end-to-end supervised learning method is proposed for fault detection in the electric grid using Big Data from multiple Phasor Measurement Units (PMUs). The approach consists of preprocessing steps aimed at reducing data noise and dimensionality, followed by utilization of six classification models considered for detecting faults. Three of the models were variants of Convolutional Neural Network (CNN) architectures that consider a single type of measurement (voltage, current or frequency) at all PMUs or all types together also at all PMUs. CNN based models were compared to traditional methods of Logistic Regression (LR), Multi-layer Perceptron (MLP) and Support Vector Machine (SVM). Evaluation was conducted on two-year data measured by PMUs at 37 locations in a large electric grid. Here, the response variable for classification were extracted from the grid-wide outage event log. Experiments show that CNN-based models outperformed traditional methods on one year out-of-sample outage detection over the entire grid.

42 ENGINEERING↗
DOE OSTI2015

On approaches to analyze the sensitivity of simulated hydrologic fluxes to model parameters in the community land model

Here, effective sensitivity analysis approaches are needed to identify important parameters or factors and their uncertainties in complex Earth system models composed of multi-phase multi-component phenomena and multiple biogeophysical-biogeochemical processes. In this study, the impacts of 10 hydrologic parameters in the Community Land Model on simulations of runoff and latent heat flux are evaluated using data from a watershed. Different metrics, including residual statistics, the Nash-Sutcliffe coefficient, and log mean square error, are used as alternative measures of the deviations between the simulated and field observed values. Four sensitivity analysis (SA) approaches, including analysis of variance based on the generalized linear model, generalized cross validation based on the multivariate adaptive regression splines model, standardized regression coefficients based on a linear regression model, and analysis of variance based on support vector machine, are investigated. Results suggest that these approaches show consistent measurement of the impacts of major hydrologic parameters on response variables, but with differences in the relative contributions, particularly for the secondary parameters. The convergence behaviors of the SA with respect to the number of sampling points are also examined with different combinations of input parameter sets and output response variables and their alternative metrics. This study helps identify the optimal SA approach, provides guidance for the calibration of the Community Land Model parameters to improve the model simulations of land surface fluxes, and approximates the magnitudes to be adjusted in the parameter values during parametric model optimization.

58 GEOSCIENCES↗
DOE OSTI2020

Machine learning-based prediction of enzyme substrate scope: Application to bacterial nitrilases

Predicting the range of substrates accepted by an enzyme from its amino acid sequence is challenging. Although sequenc- and structure-based annotation approaches are often accurate for predicting broad categories of substrate specificity, they generally cannot predict which specific molecules will be accepted as substrates for a given enzyme, particularly within a class of closely related molecules. Combining targeted experimental activity data with structural modeling, ligand docking, and physicochemical properties of proteins and ligands with various machine learning models provides complementary information that can lead to accurate predictions of substrate scope for related enzymes. Here we describe such an approach that can predict the substrate scope of bacterial nitrilases, which catalyze the hydrolysis of nitrile compounds to the corresponding carboxylic acids and ammonia. Each of the four machine learning models (logistic regression, random forest, gradient-boosted decision trees, and support vector machines) performed similarly (average ROC = 0.9, average accuracy = ~82%) for predicting substrate scope for this dataset, although random forest offers some advantages. Finally, this approach is intended to be highly modular with respect to physicochemical property calculations and software used for structural modeling and docking.

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 OSTI2025

Parallel sorting algorithm classification: is manual instrumentation necessary?

Understanding parallel algorithms is crucial for accelerating scientific simulations on complex, distributed memory, high-performance computers. Modern algorithm classification approaches learn semantics directly from source code to differentiate between algorithms, however, accessing source code is not always possible. We can learn about parallel algorithms from observing their performance, as programs running the same algorithms and using the same hardware should exhibit similar performance characteristics. We present an approach to learn algorithm classes from parallel performance data directly in order to classify algorithms without access to the source code. We extend previous work to enable classifying parallel sorting algorithms using automatic instrumentation instead of requiring manual region annotations in the source code. In this work, we design and demonstrate a study for classification of parallel sorting algorithms using parallel performance data collected from automatic instrumentation, and evaluate the performance of our new methodology on classification. We leverage Caliper to collect the performance data, Thicket for our exploratory data analysis (EDA), and PyTorch and Scikit-learn to evaluate the effectiveness of random forests, support vector machines (SVMs), decision trees, neural networks, and logistic regressions on parallel performance data. Additionally, we study noise in parallel performance data, whether the removal of noise and pre-processing of the data is necessary to accurately classify parallel sorting algorithms, and determine the effectiveness of features created from performance data. In conclusion, we demonstrate classification accuracy for these five different models of up to 97.7% across four different parallel algorithm classes.

Algorithm Classification↗
DOE OSTI2013

A FIRST LOOK AT CREATING MOCK CATALOGS WITH MACHINE LEARNING TECHNIQUES

We investigate machine learning (ML) techniques for predicting the number of galaxies (N{sub gal}) that occupy a halo, given the halo's properties. These types of mappings are crucial for constructing the mock galaxy catalogs necessary for analyses of large-scale structure. The ML techniques proposed here distinguish themselves from traditional halo occupation distribution (HOD) modeling as they do not assume a prescribed relationship between halo properties and N{sub gal}. In addition, our ML approaches are only dependent on parent halo properties (like HOD methods), which are advantageous over subhalo-based approaches as identifying subhalos correctly is difficult. We test two algorithms: support vector machines (SVM) and k-nearest-neighbor (kNN) regression. We take galaxies and halos from the Millennium simulation and predict N{sub gal} by training our algorithms on the following six halo properties: number of particles, M{sub 200}, {sigma}{sub v}, v{sub max}, half-mass radius, and spin. For Millennium, our predicted N{sub gal} values have a mean-squared error (MSE) of {approx}0.16 for both SVM and kNN. Our predictions match the overall distribution of halos reasonably well and the galaxy correlation function at large scales to {approx}5%-10%. In addition, we demonstrate a feature selection algorithm to isolate the halo parameters that are most predictive, a useful technique for understanding the mapping between halo properties and N{sub gal}. Lastly, we investigate these ML-based approaches in making mock catalogs for different galaxy subpopulations (e.g., blue, red, high M{sub star}, low M{sub star}). Given its non-parametric nature as well as its powerful predictive and feature selection capabilities, ML offers an interesting alternative for creating mock catalogs.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2023

Machine Learning Classification of Molten Salt Heat Exchanger Channel Plugging using Synthetic Data

This report addresses the requirements of Milestone M3.4 AI capability to identify and predict maintenance events. Development of digital twins (DT) for molten salt reactor (MSR) components is crucial for reducing operating and maintenance costs (O&M) and ensuring commercial viability of these reactors. Our focus is on development of DT for MSR primary system heat exchanger (HX), a critical component, the fault in which can reduce operating efficiency and force reactor shutdown. We are investigating the feasibility of a conceptual DT of HX consisting of internal distributed temperature sensing with fiber optics and machine learning (ML) algorithms to detect and localize faults. To determine the optimal approach to detection and localization of channel plugging, we benchmark seven different ML models: Logistic Regression, K-Nearest Neighbors (KNN), Gaussian Naïve Bayes, Support Vector Machines (SVM), Decision Tree Classifier, Random Forest Tree Classifier, and Feed-Forward Neural Network. ML algorithms are benchmarked using synthetic HX plugging data generated with computational fluid dynamics COMSOL software, with added brown noise to represent experimental noise. We show that the best performance is obtained with the Decision Tree classifier.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2014

Approximate l-Fold Cross-Validation with Least Squares SVM and Kernel Ridge Regression

Kernel methods have difficulties scaling to large modern data sets. The scalability issues are based on computational and memory requirements for working with a large matrix. These requirements have been addressed over the years by using low-rank kernel approximations or by improving the solvers' scalability. However, Least Squares Support Vector Machines (LS-SVM), a popular SVM variant, and Kernel Ridge Regression still have several scalability issues. In particular, the O(n^3) computational complexity for solving a single model, and the overall computational complexity associated with tuning hyper parameters are still major problems. We address these problems by introducing an O(nlog n) approximate l-fold cross-validation method that uses a multi-level circulant matrix to approximate the kernel. In addition, we prove our algorithm's computational complexity and present empirical runtimes on data sets with approximately one million data points. We also validate our approximate method's effectiveness at selecting hyper parameters on real world and standard benchmark data sets. Lastly, we provide experimental results on using a multi level circulant kernel approximation to solve LS-SVM problems with hyper parameters selected using our method.

Computer Science↗
DOE OSTI2018

Integrating Natural Language Processing and Machine Learning Algorithms to Categorize Oncologic Response in Radiology Reports

A significant volume of medical data remains unstructured. Natural language processing (NLP) and machine learning (ML) techniques have shown to successfully extract insights from radiology reports. However, the codependent effects of NLP and ML in this context have not been well-studied. Between April 1, 2015 and November 1, 2016, 9418 cross-sectional abdomen/pelvis CT and MR examinations containing our internal structured reporting element for cancer were separated into four categories: Progression, Stable Disease, Improvement, or No Cancer. We combined each of three NLP techniques with five ML algorithms to predict the assigned label using the unstructured report text and compared the performance of each combination. The three NLP algorithms included term frequency-inverse document frequency (TF-IDF), term frequency weighting (TF), and 16-bit feature hashing. The ML algorithms included logistic regression (LR), random decision forest (RDF), one-vs-all support vector machine (SVM), one-vs-all Bayes point machine (BPM), and fully connected neural network (NN). The best-performing NLP model consisted of tokenized unigrams and bigrams with TF-IDF. Increasing N-gram length yielded little to no added benefit for most ML algorithms. With all parameters optimized, SVM had the best performance on the test dataset, with 90.6 average accuracy and F score of 0.813. The interplay between ML and NLP algorithms and their effect on interpretation accuracy is complex. The best accuracy is achieved when both algorithms are optimized concurrently.

62 RADIOLOGY AND NUCLEAR MEDICINE↗