Search NASA⌕ Search

SEARCH · Search NASA

Results for “variable importance”

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 19 records

Modeling freight mode choice using machine learning classifiers: a comparative study using Commodity Flow Survey (CFS) data

This study explores the usefulness of machine learning classifiers for modeling freight mode choice. We investigate eight commonly used machine learning classifiers, namely Naïve Bayes, Support Vector Machine, Artificial Neural Network, K-Nearest Neighbors, Classification and Regression Tree, Random Forest, Boosting and Bagging, along with the classical Multinomial Logit model. US 2012 Commodity Flow Survey data are used as the primary data source; we augment it with spatial attributes from secondary data sources. The performance of the classifiers is compared based on prediction accuracy results. The current research also examines the role of sample size and training-testing data split ratios on the predictive ability of the various approaches. In addition, the importance of variables is estimated to determine how the variables influence freight mode choice. The results show that the tree-based ensemble classifiers perform the best. Specifically, Random Forest produces the most accurate predictions, closely followed by Boosting and Bagging. With regard to variable importance, shipment characteristics, such as shipment distance, industry classification of the shipper and shipment size, are the most significant factors for freight mode choice decisions.

42 ENGINEERING↗

An interpretable machine learning model for advancing terrestrial ecosystem predictions

We apply an interpretable Long Short-Term Memory (iLSTM) network for land-atmosphere carbon flux predictions based on time series observations of seven environmental variables. iLSTM enables interpretability of variable importance and variable-wise temporal importance to the prediction of targets by exploring internal network structures. The application results indicate that iLSTM not only improves prediction performance by capturing different dynamics of individual variables, but also reasonably interprets the different contribution of each variable to the target and its different temporal relevance to the target. This variable and temporal importance interpretation of iLSTM advances terrestrial ecosystem model development as well as our predictive understanding of the system.

Lu, Dan↗

Investigation of hydrometeorological influences on reservoir releases using explainable machine learning methods

Long short-term memory (LSTM) networks have demonstrated successful applications in accurately and efficiently predicting reservoir releases from hydrometeorological drivers including reservoir storage, inflow, precipitation, and temperature. However, due to its black-box nature and lack of process-based implementation, we are unsure whether LSTM makes good predictions for the right reasons. In this work, we use an explainable machine learning (ML) method, called SHapley Additive exPlanations (SHAP), to evaluate the variable importance and variable-wise temporal importance in the LSTM model prediction. In application to 30 reservoirs over the Upper Colorado River Basin, United States, we show that LSTM can accurately predict the reservoir releases with NSE ≥ 0.69 for all the considered reservoirs despite of their diverse storage sizes, functionality, elevations, etc. Additionally, SHAP indicates that storage and inflow are more influential than precipitation and temperature. Moreover, the storage and inflow show a relatively long-term influence on the release up to 7 days and this influence decreases as the lag time increases for most reservoirs. These findings from SHAP are consistent with our physical understanding. However, in a few reservoirs, SHAP gives some temporal importances that are difficult to interpret from a hydrological point of view, probably because of its ignorance of the variable interactions. SHAP is a useful tool for black-box ML model explanations, but the hydrological processes inferred from its results should be interpreted cautiously. More investigations of SHAP and its applications in hydrological modeling is needed and will be pursued in our future study.

54 ENVIRONMENTAL SCIENCES↗

Spatial Heterogeneity in Ecologically Important Climate Variables at Coarse and Fine Scales in a High-Snow Mountain Landscape

Climate plays an important role in determining the geographic ranges of species. With rapid climate change expected in the coming decades, ecologists have predicted that species ranges will shift large distances in elevation and latitude. However, most range shift assessments are based on coarse-scale climate models that ignore fine-scale heterogeneity and could fail to capture important range shift dynamics. Moreover, if climate varies dramatically over short distances, some populations of certain species may only need to migrate tens of meters between microhabitats to track their climate as opposed to hundreds of meters upward or hundreds of kilometers poleward. To address these issues, we measured climate variables that are likely important determinants of plant species distributions and abundances (snow disappearance date and soil temperature) at coarse and fine scales at Mount Rainier National Park in Washington State, USA. Coarse-scale differences across the landscape such as large changes in elevation had expected effects on climatic variables, with later snow disappearance dates and lower temperatures at higher elevations. However, locations separated by small distances (~20 m), but differing by vegetation structure or topographic position, often experienced differences in snow disappearance date and soil temperature as great as locations separated by large distances (>1 km). Tree canopy gaps and topographic depressions experienced later snow disappearance dates than corresponding locations under intact canopy and on ridges. Additionally, locations under vegetation and on topographic ridges experienced lower maximum and higher minimum soil temperatures. The large differences in climate we observed over small distances will likely lead to complex range shift dynamics and could buffer species from the negative effects of climate change.

54 ENVIRONMENTAL SCIENCES↗

Influence of Vegetation and Seasonal Forcing on Carbon Dioxide Fluxes Across the Upper Midwest, USA: Implications for Regional Scaling

Carbon dioxide fluxes were examined over the growing seasons of 2002 and 2003 from 14 different sites in the Upper Midwest (USA) to assess spatial variability of ecosystem–atmosphere CO2 exchange. These sites were exposed to similar temperature/precipitation regimes and spanned a range of vegetation types typical of the region (northern hardwood, mixed forest, red pine, jack pine, pine barrens, and shrub wetland). The hardwood and red pine sites also spanned a range of stand ages (young, intermediate, mature). While seasonal changes in net ecosystem exchange (NEE) and photosynthetic parameters were coherent across the 2 years at most sites, changes in ecosystem respiration (ER) and gross ecosystem production (GEP) were not. Canopy height and vegetation type were important variables for explaining spatial variability of CO2 fluxes across the region. Light-use efficiency (LUE) was not as strongly correlated to GEP as maximum assimilation capacity (Amax). A bottom-up multi-tower land cover aggregated scaling of CO2 flux to a 2000 km2 regional flux estimate found June to August 2003 NEE, ER, and GEP to be -290 ± 89, 408 ± 48, and 698 ± 73 gC m -2 , respectively. Aggregated NEE, ER, and GEP were 280% larger, 32% smaller and 3% larger, respectively, than that observed from a regionally integrating 447 m tall flux tower. However, when the tall tower fluxes were decomposed using a footprint-weighted influence function and then re-aggregated to a regional estimate, the resulting NEE, ER, and GEP were within 11% of the multi-tower aggregation. Excluding wetland and young stand age sites from the aggregation worsened the comparison to observed fluxes. These results provide insight on the range of spatial sampling, replication, measurement error, and land cover accuracy needed for multi-tiered bottom-up scaling of CO2 fluxes in heterogeneous regions such as the Upper Midwest, USA.

54 ENVIRONMENTAL SCIENCES↗

Classification Analysis of Southwest Pacific Tropical Cyclone Intensity Changes Prior to Landfall

This study evaluates the ability of a random forest classifier to identify tropical cyclone (TC) intensification or weakening prior to landfall over the western region of the Southwest Pacific Ocean (SWPO) basin. For both Australia mainland and SWPO island cases, when a TC first crosses land after spending ≥24 h over the ocean, the closest hour prior to the intersection is considered as the landfall hour. If the maximum wind speed (V max ) at the landfall hour increased or remained the same from the 24-h mark prior to landfall, the TC is labeled as intensifying and if the V max at the landfall hour decreases, the TC is labeled as weakening. Geophysical and aerosol variables closest to the 24 h before landfall hour were collected for each sample. The random forest model with leave-one-out cross validation and the random oversampling example technique was identified as the best-performing classifier for both mainland and island cases. The model identified longitude, initial intensity, and sea skin temperature as the most important variables for the mainland and island landfall classification decisions. Incorrectly classified cases from the test data were analyzed by sorting the cases by their initial intensity hour, landfall hour, monthly distribution, and 24-h intensity changes. TC intensity changes near land strongly impact coastal preparations such as wind damage and flood damage mitigations; hence, this study will contribute to improve identifying and prioritizing prediction of important variables contributing to TC intensity change before landfall.

54 ENVIRONMENTAL SCIENCES↗

Southwest Pacific tropical cyclone development classification utilizing machine learning and synoptic composites

This study evaluates the ability of machine learning algorithms to classify tropical depressions (TDs) and tropical storms (TSs) in the western region of the southwest Pacific Ocean (SWPO). Decision rules are generated to predict the environment required for a depression to fully develop into a mature storm, and the most influential predictors in the classification decision are ranked. TD and TS are discriminated based on a maximum sustained wind speed threshold (≥17 ms -1 ). Various aerosol, thermodynamic, and dynamic parameters are extracted closest to the initiation point of each non-developing and developing sample. The covariates associated with each labelled sample are used to train a decision tree and random forest model. Results using a testing dataset suggest the random forest approach more accurately distinguishes between non-developing and developing samples. The classification accuracy of the decision tree and random forest are 72% and 91%, respectively. Random forest outperformed the decision tree by providing higher accuracy in test data. The most important variables for binary classification are sea salt aerosol optical depth (AOD), 1,000 mb relative humidity, and sea surface temperature. AOD is a quantitative estimate of the aerosols presents in the air through the extinction of a ray of light as it passes through the atmosphere. Mean composite maps constructed in an unsupervised manner have been created for the most important variables identified by the random forest classifier during TD and TS events to highlight the difference in geophysical and aerosol variables' climatology during the two different classifications. This work will advance the risk management strategies for northeastern Australia and other SWPO basin islands to control their tropical cyclone related losses through prioritizing forecasting variables that are the strongest predictors of the strengthening of tropical depressions into tropical cyclones.

54 ENVIRONMENTAL SCIENCES↗

Small field in-air output factors: The role of miniphantom design and dosimeter type

Purpose: The commissioning of treatment planning systems and beam modeling requires measured input parameters. The measurement of relative output in-air, S{sub c} is particularly difficult for small fields. The purpose of this study was to investigate the influence of miniphantom design and detector selection on measured S{sub c} values for small fields and to validate the measurements against Monte Carlo simulations. Methods: Measurements were performed using brass caps (with sidewalls) or tops (no sidewalls) of varying heights and widths. The performance of two unshielded diodes (60012 and SFD), EBT2 radiochromic film, and a fiber optic dosimeter (FOD) were compared for fields defined by MLCs (5–100 mm) and SRS cones (4–30 mm) on a Varian Novalis linear accelerator. Monte Carlo simulations were performed to theoretically predict S{sub c} as measured by the FOD. Results: For all detectors, S{sub c} agreed to within 1% for fields larger than 10 mm and to within 2.3% for smaller fields. Monte Carlo simulation matched the FOD measurements for all size of cone defined fields to within 0.5%. Conclusions: Miniphantom design is the most important variable for reproducible and accurate measurements of the in-air output ratio, S{sub c}, in small photon fields (less than 30 mm). Sidewalls are not required for fields ≤ 30 mm and tops are therefore preferred over the larger caps. Unlike output measurements in water, S{sub cp,} the selection of detector type for S{sub c} is not critical, provided the active dosimeter volume is small relative to the field size.

61 RADIATION PROTECTION AND DOSIMETRY↗

Implementation of a feature selection algorithm in FARM to identify important state variables and time-invariant matrices

The FARM (Feasible Actuator Range Modifier) software module is a component of the RAVEN-based FORCE framework for analysis of Integrated Energy Systems (IES). FARM aids the HERON software module in the evaluation of the optimal dispatch for the different IES components. Set-point trajectories are required to meet limits on both production variables (i.e., the variables to be optimized such as the electrical power, the hydrogen production rate, etc.) and process variables tied to the service life of equipment (e.g., steam flowrate, vessel pressure, turbine firing temperature, etc.). To evaluate the feasibility of HERON generated set-points and to do so in an acceptable time, FARM employs reduced order models to represent the dynamic behavior of the systems to be dispatched. These surrogate models take the form of a linear dynamic system with sets of Linear Parameter Varying (LPV) matrices that are mapped to the system operating space. These matrices are derived from the trajectories of system state variables and system output variables during transients. The accuracy of LPV matrices depends on the selection of state variables. In previous reports, state variables were selected by adopting a complicated workflow requiring multiple software licenses and an advanced level of user expertise. In this report, a new workflow that automates the state variable selection process is presented. It significantly reduces the frequency of user interventions and does not require multiple software licenses. Each module in the new workflow is described in detail, and the input / output examples in each step of the workflow are provided. It was demonstrated that this workflow can greatly reduce the complexity of the state variable selection process, and that the updated FARM-Gamma and FARM-Delta validators can benefit from this workflow when solving the power dispatch problem of a representative IES test case. Finally, some code improvements that can further enhance the efficiency are suggested.

42 ENGINEERING↗

Machine Learning in Infectious Disease for Risk Factor Identification and Hypothesis Generation: Proof of Concept Using Invasive Candidiasis

Machine learning (ML) models can handle large data sets without assuming underlying relationships and can be useful for evaluating disease characteristics, yet they are more commonly used for predicting individual disease risk than for identifying factors at the population level. We offer a proof of concept applying random forest (RF) algorithms to Candida-positive hospital encounters in an electronic health record database of patients in the United States. Candida-positive encounters were extracted from the Cerner HealthFacts database; invasive infections were laboratory-positive sterile site Candida infections. Features included demographics, admission source, care setting, physician specialty, diagnostic and procedure codes, and medications received before the first positive Candida culture. We used RF to assess risk factors for 3 outcomes: any invasive candidiasis (IC) vs non-IC, within-species IC vs non-IC (eg, invasive C. glabrata vs noninvasive C. glabrata), and between-species IC (eg, invasive C. glabrata vs all other IC). Fourteen of 169 (8%) variables were consistently identified as important features in the ML models. When evaluating within-species IC, for example, invasive C. glabrata vs non-invasive C. glabrata, we identified known features like central venous catheters, intensive care unit stay, and gastrointestinal operations. In contrast, important variables for invasive C. glabrata vs all other IC included renal disease and medications like diabetes therapeutics, cholesterol medications, and antiarrhythmics. Known and novel risk factors for IC were identified using ML, demonstrating the hypothesis-generating utility of this approach for infectious disease conditions about which less is known, specifically at the species level or for rarer diseases.

60 APPLIED LIFE SCIENCES↗

Identifying Entangled Physics Relationships Through Sparse Matrix Decomposition to Inform Plasma Fusion Design

We report a sustainable burn platform through inertial confinement fusion (ICF) has been an ongoing challenge for over 50 years. Mitigating engineering limitations and improving the current design involves an understanding of the complex coupling of physical processes. While sophisticated simulation codes are used to model ICF implosions, these tools contain necessary numerical approximation but miss physical processes that limit predictive capability. Identification of relationships between controllable design inputs to ICF experiments and measurable outcomes (e.g., neutron yield, neutron velocity, areal density) from performed experiments can help guide the future design of experiments and development of simulation codes, to potentially improve the accuracy of the computational models used to simulate ICF experiments. We use sparse matrix decomposition methods to identify clusters of a few related design variables. Sparse principal component analysis (SPCA) identifies groupings that are related to the physical origin of the variables (laser, hohlraum, and capsule). A variable importance analysis finds that in addition to variables highly correlated with neutron yield, such as picket power and laser energy, variables that represent a dramatic change of the ICF design, such as number of pulse steps, are also very important. The obtained sparse components are then used to train a random forest (RF) regression surrogate for predicting total yield. The RF performance on the training and testing data compares with the performance of the RF trained using all the design variables considered. This work is intended to inform design changes in future ICF experiments by augmenting the expert intuition and simulation results.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Data‐driven identification of environmental variables influencing phenotypic plasticity to facilitate breeding for future climates

Summary Phenotypic plasticity describes a genotype's ability to produce different phenotypes in response to different environments. Breeding crops that exhibit appropriate levels of plasticity for future climates will be crucial to meeting global demand, but knowledge of the critical environmental factors is limited to a handful of well‐studied major crops. Using 727 maize ( Zea mays L.) hybrids phenotyped for grain yield in 45 environments, we investigated the ability of a genetic algorithm and two other methods to identify environmental determinants of grain yield from a large set of candidate environmental variables constructed using minimal assumptions. The genetic algorithm identified pre‐ and postanthesis maximum temperature, mid‐season solar radiation, and whole season net evapotranspiration as the four most important variables from a candidate set of 9150. Importantly, these four variables are supported by previous literature. After calculating reaction norms for each environmental variable, candidate genes were identified and gene annotations investigated to demonstrate how this method can generate insights into phenotypic plasticity. The genetic algorithm successfully identified known environmental determinants of hybrid maize grain yield. This demonstrates that the methodology could be applied to other less well‐studied phenotypes and crops to improve understanding of phenotypic plasticity and facilitate breeding crops for future climates.

Kusmec, Aaron↗

Ten-year variability in ecosystem water use efficiency in an oak-dominated temperate forest under a warming climate

The impacts of extreme weather events on water-carbon (C) coupling and ecosystem-scale water use efficiency (WUE) over a long term are poorly understood. We analyzed the changes in ecosystem water use efficiency (WUE) from 10 years of eddy-covariance measurements (2004-2013) over an oak-dominated temperate forest in Ohio, USA. The aim was to investigate the long-term response of ecosystem WUE to measured changes in site-biophysical conditions and ecosystem attributes. The oak forest produced new plant biomass of 2.5 +/- 0.2 gC kg(-1) of water loss annually. Monthly evapotranspiration (ET) and gross ecosystem production (GEP) were tightly coupled over the 10-year study period (R-2=0.94). Daily WUE had a linear relationship with air temperature (T-a) in low-temperature months and a unimodal relationship with T-a in high-temperature months during the growing season. On average, daily WUE ceased to increase when T-a exceeded 22 degrees C in warm months for both wet and dry years. Monthly WUE had a strong positive linear relationship with leaf area index (LAI), net radiation (R-n), and T-a and weak logarithmic relationship with water vapor pressure deficit (VPD) and precipitation (P) on a growing-season basis. When exploring the regulatory mechanisms on WUE within each season, spring LAI and P, summer R-n and T-a, and autumnal VPD and R-n were found to be the main explanatory variables for seasonal variation in WUE. The model developed in this study was able to capture 78% of growing-season variation in WUE on a monthly basis. The negative correlation between WUE and A in spring was mainly due to the high precipitation amounts in spring, decreasing GEP and WUE when LAI was still small, adding ET being observed to increase with high levels of evaporation as a result of high SWC in spring. Summer WUE had a significant decreasing trend across the 10 years mainly due to the combined effect of seasonal drought and increasing potential and available energy increasing ET, but decreasing GEP in summer. We concluded that seasonal dynamics of the interchange between precipitation and drought status of the system was an important variable in controlling seasonal WUE in wet years. In contrast, despite the negative impacts of unfavorable warming, available groundwater and an early start of the growing season were important contributing variables in high seasonal GEP, and thus, high seasonal WUE in dry years. (C) 2015 Elsevier B.V. All rights reserved.

54 ENVIRONMENTAL SCIENCES↗

A new variable-resolution associative memory for high energy physics

We describe an important advancement for the Associative Memory device (AM). The AM is a VLSI processor for pattern recognition based on Content Addressable Memory (CAM) architecture. The AM is optimized for on-line track finding in high-energy physics experiments. Pattern matching is carried out by finding track candidates in coarse resolution 'roads'. A large AM bank stores all trajectories of interest, called 'patterns', for a given detector resolution. The AM extracts roads compatible with a given event during detector read-out. Two important variables characterize the quality of the AM bank: its 'coverage' and the level of fake roads. The coverage, which describes the geometric efficiency of a bank, is defined as the fraction of tracks that match at least one pattern in the bank. Given a certain road size, the coverage of the bank can be increased just adding patterns to the bank, while the number of fakes unfortunately is roughly proportional to the number of patterns in the bank. Moreover, as the luminosity increases, the fake rate increases rapidly because of the increased silicon occupancy. To counter that, we must reduce the width of our roads. If we decrease the road width using the current technology, the system will become very large and extremely expensive. We propose an elegant solution to this problem: the 'variable resolution patterns'. Each pattern and each detector layer within a pattern will be able to use the optimal width, but we will use a 'don't care' feature (inspired from ternary CAMs) to increase the width when that is more appropriate. In other words we can use patterns of variable shape. As a result we reduce the number of fake roads, while keeping the efficiency high and avoiding excessive bank size due to the reduced width. We describe the idea, the implementation in the new AM design and the implementation of the algorithm in the simulation. Finally we show the effectiveness of the 'variable resolution patterns' idea using simulated high occupancy events in the ATLAS detector. (authors)

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Inter-well connectivity detection in CO 2 WAG projects using statistical recurrent unit models

Routine well-wise injection and production measurements contain significant information on subsurface structure and properties. Data-driven technology that interprets surface data into subsurface structure or properties can assist operators in making informed decisions by providing a better understanding of field assets. Our machine-learning framework is built on the statistical recurrent unit (SRU) model and interprets well-based injection/production data into inter-well connectivity without relying on a geologic model. We test it on synthetic and field-scale CO 2 EOR projects utilizing the water-alternating-gas (WAG) process. SRU is a special type of recurrent neural network (RNN) that allows for better characterization of temporal trends, by learning various statistics of the input at different time scales. In our application, the complete states (injection rate, pressure and cumulative injection) at injectors and pressure states at producers are fed to SRU as the input and the phase rates at producers are treated as the output. Once the SRU is trained and validated, it is then used to assess the connectivity of each injector to any producer using permutation variable importance method, wherein inputs corresponding to an injector are shuffled and the increase in prediction error at a given producer is recorded as the importance (connectivity metric) of the injector to the producer. This method is tested in both synthetic and field-scale cases. The validation of the proposed data-driven inter-well connectivity assessment is performed using synthetic data from simulation models where inter-well connectivity can be easily measured using the streamline-based flux allocation. The SRU model is shown to offer excellent prediction performance on the synthetic case. Despite significant measurement noise and frequent well shut-ins imposed in the field-scale case, the SRU model offers good prediction accuracy, the overall relative error of the phase production rates at most producers ranges from 10% to 30%. It is shown that the dominant connections identified by the data-driven method and streamline method are in close agreement. This significantly improves confidence in our data-driven procedure. The novelty of this work is that it is purely data-driven method and can directly interpret routine surface measurements to intuitive subsurface knowledge. Furthermore, the streamline-based validation procedure provides physics-based backing to the results obtained from data analytics. This study results in a reliable and efficient data analytics framework that is well-suited for large field applications.

42 ENGINEERING↗

Stepwise v. 2.21

The statistical code, STEPWISE, evaluates variable importance by developing regression models between the observed response and input variables using either a forward, backward, or stepwise regression procedure on the raw or ranked data. The software will be used for performing sensitivity analysis on probabilistic results. Stepwise creates linear (or monotonic) regression of the output of interest using a set of input data in a stepwise fashion (i.e., adding one variable at a time). The program estimates parameters such as regression and standardized regressions coefficients for each selected input variables as well as the coefficient of determination of the resulting model

Sallaberry, CedricJean-Marie↗

Evaluating geothermal and hydrogeologic controls on regional groundwater temperature distribution

Described is a one-dimensional (1-D) analytic solution is developed for heat transport through an aquifer system where the vertical temperature profile in the aquifer is nearly uniform. The general anisotropic form of the viscous heat generation term is developed for use in groundwater flow simulations. The 1-D solution is extended to more complex geometries by solving the equation for piece-wise linear or uniform properties and boundary conditions. A moderately complex example, the Eastern Snake River Plain (ESRP), is analyzed to demonstrate the use of the analytic solution for identifying important physical processes. For example, it is shown that viscous heating is variably important and that heat conduction to the land surface is a primary control on the distribution of aquifer and spring temperatures. Use of published values for all aquifer and thermal properties results in a reasonable match between simulated and measured groundwater temperatures over most of the 300 km length of the ESRP, except for geothermal heat flow into the base of the aquifer within 20 km of the Yellowstone hotspot. Previous basal heat flow measurements (110 mW/m 2 ) made beneath the ESRP aquifer were collected at distances of > 50 km from the Yellowstone Plateau, but a higher basal heat flow of 150 mW/m 2 is required to match groundwater temperatures near the Plateau. As a result, the ESRP example demonstrates how the new tool can be used during preliminary analysis of a groundwater system, allowing efficient identification of the important physical processes that must be represented during more-complex 2-D and 3-D simulations of combined groundwater and heat flow.

58 GEOSCIENCES↗