SEARCH · Search NASA
Results for “Shapley value”
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.
Explainable AI for a No-Teardown Vehicle Component Cost Estimation: A Top-Down Approach
The broader ambition of this article is to popularize an approach for the fair distribution of the quantity of a system's output to its subsystems while allowing for underlying complex subsystem level interactions. Particularly, we present the use of this framework on a very specific (but generalizable) application, interesting for a more general AI audience. Here, we detail a data-driven approach to vehicle price modeling and its component price estimation by leveraging a combination of concepts from machine learning and game theory. We show an alternative to common teardown methodologies and surveying approaches for component and vehicle price estimation at the manufacturer's suggested retail price (MSRP) level that has the advantage of bypassing uncertainties involved in gathering teardown data, the need to perform expensive and biased surveying, and the need to perform retail price equivalent or indirect cost multiplier adjustments to mark up direct manufacturing costs to MSRP. This novel exercise not only provides accurate pricing of the technologies at the customer level, but also shows the, a priori known, large gaps in pricing strategies between manufacturers, vehicle classes, market segments, etc. There is also clear interaction between the price of technologies and other specifications present in vehicles. Those results are indication that old methods of manufacturer-level component costing, aggregation, and application of flat and rigid adjustment factors should be carefully examined. The findings are based on a database developed by Argonne, which includes over 64,000 vehicles covering MY1990 to MY2020 with hundreds of vehicle specs.
Extracting structural motifs from pair distribution function data of nanostructures using explainable machine learning
Characterization of material structure with X-ray or neutron scattering using e.g. Pair Distribution Function (PDF) analysis most often rely on refining a structure model against an experimental dataset. However, identifying a suitable model is often a bottleneck. Recently, automated approaches have made it possible to test thousands of models for each dataset, but these methods are computationally expensive and analysing the output, i.e. extracting structural information from the resulting fits in a meaningful way, is challenging. Our Machine Learning based Motif Extractor (ML-MotEx) trains an ML algorithm on thousands of fits, and uses SHAP (SHapley Additive exPlanation) values to identify which model features are important for the fit quality. We use the method for 4 different chemical systems, including disordered nanomaterials and clusters. ML-MotEx opens for a type of modelling where each feature in a model is assigned an importance value for the fit quality based on explainable ML.
An Investigation of LES Wall Modeling for Rayleigh–Bénard Convection via Interpretable and Physics-Aware Feedforward Neural Networks with DNS
Abstract The traditional approach of using the Monin–Obukhov similarity theory (MOST) to model near-surface processes in large-eddy simulations (LESs) can lead to significant errors in natural convection. In this study, we propose an alternative approach based on feedforward neural networks (FNNs) trained on output from direct numerical simulation (DNS). To evaluate the performance, we conduct both a priori and a posteriori tests. In the a priori (offline) tests, we compare the statistics of the surface shear stress and heat flux, computed from filtered DNS input variables, to the stress and flux obtained from the filtered DNS. Additionally, we investigate the importance of various input features using the Shapley additive explanations value and the conditional average of the filter grid cells. In the a posteriori (online) tests, we implement the trained models in the System for Atmospheric Modeling (SAM) LES and compare the LES-generated surface shear stress and heat flux with those in the DNS. Our findings reveal that vertical velocity, a traditionally overlooked flow quantity, is one of the most important input features for determining the wall fluxes. Increasing the number of input features improves the a priori test results but does not always improve the model performance in the a posteriori tests because of the differences in input variables between the LES and DNS. Last, we show that physics-aware FNN models trained with logarithmic and scaled parameters can well extrapolate to more intense convection scenarios than in the training dataset, whereas those trained with primitive flow quantities cannot. Significance Statement The traditional near-surface turbulence model, based on a shear-dominated boundary layer flow, does not represent near-surface turbulence in natural convection. Using a feedforward neural network (FNN), we can construct a more accurate model that better represents the near-surface turbulence in various flows and reveals previously overlooked controlling factors and process interactions. Our study shows that the FNN-generated models outperform the traditional model and highlight the importance of the near-surface vertical velocity. Furthermore, the physics-aware FNN models exhibit the potential to extrapolate to convective flows of various intensities beyond the range of the training dataset, suggesting their broader applicability for more accurate modeling of near-surface turbulence.
Exploring the dependence of gas cooling and heating functions on the incident radiation field with machine learning
ABSTRACT Gas cooling and heating functions play a crucial role in galaxy formation. But, it is computationally expensive to exactly compute these functions in the presence of an incident radiation field. These computations can be greatly sped up by using interpolation tables of pre-computed values, at the expense of making significant and sometimes even unjustified approximations. Here, we explore the capacity of machine learning to approximate cooling and heating functions with a generalized radiation field. Specifically, we use the machine learning algorithm XGBoost to predict cooling and heating functions calculated with the photoionization code cloudy at fixed metallicity, using different combinations of photoionization rates as features. We perform a constrained quadratic fit in metallicity to enable a fair comparison with traditional interpolation methods at arbitrary metallicity. We consider the relative importance of various photoionization rates through both a principal component analysis (PCA) and calculation of SHapley Additive exPlanation (shap) values for our XGBoost models. We use feature importance information to select different subsets of rates to use in model training. Our XGBoost models outperform a traditional interpolation approach at each fixed metallicity, regardless of feature selection. At arbitrary metallicity, we are able to reduce the frequency of the largest cooling and heating function errors compared to an interpolation table. We find that the primary bottleneck to increasing accuracy lies in accurately capturing the metallicity dependence. This study demonstrates the potential of machine learning methods such as XGBoost to capture the non-linear behaviour of cooling and heating functions.
Macroscopic Traffic Modeling Using Probe Vehicle Data: A Machine Learning Approach
Abstract The macroscopic fundamental diagram (MFD) captures an orderly relationship among traffic flow, density, and speed at the network level. It is a simple yet powerful tool for modeling traffic dynamics in large urban networks with broad application in traffic control and management. However, empirically derived MFDs in urban regions require high-resolution traffic data from the network. Having the network flow and vehicular density estimated at the (granular) census tract level using vehicle probe data, we apply machine learning methods to predict the MFDs across U.S. urban areas and capture the impacts of location-specific input features on the network flow–density relationships at a large scale. The results show that, among the four tested machine learning approaches (Random Forest, XGBoost, Support Vector Machine, and Neural Network), XGBoost delivers the best performance in predicting network traffic flow based on vehicular density and location attributes. Using interaction Shapley Additive explanation (SHAP) values and partial correlation analysis, we examine the factors influencing MFD shapes across different locations. Our empirical findings reveal that across U.S. urban areas, network topology, transportation infrastructure, and land use are primary factors shaping MFD curves, while demand and trip-related factors play a lesser role. Specifically, higher ranking roads, centrality, and development levels correlate positively with network capacity and critical density, whereas negative associations are observed for network connectivity, mixed-use development, and road roughness levels.
On the minimum number of radiation field parameters to specify gas cooling and heating functions
Fast and accurate approximations of gas cooling and heating functions are needed for hydrodynamic galaxy simulations. We use machine learning to analyze atomic gas cooling and heating functions in the presence of a generalized incident local radiation field computed by Cloudy. We characterize the radiation field through binned radiation field intensities instead of the photoionization rates used in our previous work. We find a set of 6 energy bins whose intensities exhibit relatively low correlation. We use these bins as features to train machine learning models to predict Cloudy cooling and heating functions at fixed metallicity. We compare the relative SHapley Additive exPlanation (SHAP) value importance of the features. From the SHAP analysis, we identify a feature subset of 3 energy bins (0.5-1, 1-4, and 13-16Ry) with the largest importance and train additional models on this subset. We compare the mean squared errors and distribution of errors on both the entire training data table and a randomly selected 20% test set withheld from model training. The machine learning models trained with 3 and 6 bins, as well as 3 and 4 photoionization rates, have comparable accuracy everywhere, with errors ≳10 times smaller than for the interpolation table of Gnedin and Hollon (2012). We conclude that 3 energy bins (or 3 analogous photoionization rates: molecular hydrogen photodissociation, neutral hydrogen HI, and fully ionized carbon CVI) are sufficient to characterize the dependence of the gas cooling and heating functions on our assumed incident radiation field model.
Machine-learning based approach to examine ecological processes influencing the diversity of riverine dissolved organic matter composition
Dissolved organic matter (DOM) assemblages in freshwater rivers are formed from mixtures of simple to complex compounds that are highly variable across time and space. These mixtures largely form due to the environmental heterogeneity of river networks and the contribution of diverse allochthonous and autochthonous DOM sources. Most studies are, however, confined to local and regional scales, which precludes an understanding of how these mixtures arise at large, e.g., continental, spatial scales. The processes contributing to these mixtures are also difficult to study because of the complex interactions between various environmental factors and DOM. Here we propose the use of machine learning (ML) approaches to identify ecological processes contributing toward mixtures of DOM at a continental-scale. We related a dataset that characterized the molecular composition of DOM from river water and sediment with Fourier-transform ion cyclotron resonance mass spectrometry to explanatory physicochemical variables such as nutrient concentrations and stable water isotopes ( 2 H and 18 O). Using unsupervised ML, distinctive clusters for sediment and water samples were identified, with unique molecular compositions influenced by environmental factors like terrestrial input and microbial activity. Sediment clusters showed a higher proportion of protein-like and unclassified compounds than water clusters, while water clusters exhibited a more diversified chemical composition. We then applied a supervised ML approach, involving a two-stage use of SHapley Additive exPlanations (SHAP) values. In the first stage, SHAP values were obtained and used to identify key physicochemical variables. These parameters were employed to train models using both the default and subsequently tuned hyperparameters of the Histogram-based Gradient Boosting (HGB) algorithm. The supervised ML approach, using HGB and SHAP values, highlighted complex relationships between environmental factors and DOM diversity, in particular the existence of dams upstream, precipitation events, and other watershed characteristics were important in predicting higher chemical diversity in DOM. Our data-driven approach can now be used more generally to reveal the interplay between physical, chemical, and biological factors in determining the diversity of DOM in other ecosystems.
On the minimum number of radiation field parameters to specify gas cooling and heating functions
Fast and accurate approximations of gas cooling and heating functions are needed for hydrodynamic galaxy simulations. We use machine learning to analyze atomic gas cooling and heating functions computed by Cloudy in the presence of a generalized incident local radiation field. We characterize the radiation field through binned radiation field intensities instead of the photoionization rates used in our previous work. We find a set of 6 energy bins whose intensities exhibit relatively low correlation. We use these bins as features to train machine learning models to predict Cloudy cooling and heating functions at fixed metallicity. We compare the relative SHapley Additive exPlanation (SHAP) value importance of the features. From the SHAP analysis, we identify a feature subset of 3 energy bins ($0.5-1, 1-4$, and $13-16 \, \mathrm{Ry}$) with the largest importance and train additional models on this subset. We compare the mean squared errors and distribution of errors on both the entire training data table and a randomly selected 20% test set withheld from model training. The machine learning models trained with 3 and 6 bins, as well as 3 and 4 photoionization rates, have comparable accuracy everywhere, with errors $\gtrsim 10$ times smaller than for the interpolation table of Gnedin and Hollon (2012). We conclude that 3 energy bins (or 3 analogous photoionization rates: molecular hydrogen photodissociation, neutral hydrogen HI, and fully ionized carbon CVI) are sufficient to characterize the dependence of the gas cooling and heating functions on our assumed incident radiation field model.
Prognostic analysis of high-flow nasal cannula therapy and non-invasive ventilation in mild to moderate hypoxemia patients and construction of a machine learning model for 48-h intubation prediction—a retrospective analysis of the MIMIC database
Background This study aims to investigate the clinical outcome between high-flow nasal cannula (HFNC) and non-invasive ventilation (NIV) therapy in mild to moderate hypoxemic patients on the first ICU day and to develop a predictive model of 48-h intubation. Methods The study included adult patients from the MIMIC III and IV databases who first initiated HFNC or NIV therapy due to mild to moderate hypoxemia (100 < PaO2/FiO2 ≤ 300). The 48-h and 30-day intubation rates were compared using cross-sectional and survival analysis. Nine machine learning and six ensemble algorithms were deployed to construct the 48-h intubation predictive models, of which the optimal model was determined by its prediction accuracy. The top 10 risk and protective factors were identified using the Shapley interpretation algorithm. Result A total of 123,042 patients were screened, of which, 673 were from the MIMIC IV database for ventilation therapy comparison (HFNC n = 363, NIV n = 310) and 48-h intubation predictive model construction (training dataset n = 471, internal validation set n = 202) and 408 were from the MIMIC III database for external validation. The NIV group had a lower intubation rate (23.1% vs. 16.1%, p = 0.001), ICU 28-day mortality (18.5% vs. 11.6%, p = 0.014), and in-hospital mortality (19.6% vs. 11.9%, p = 0.007) compared to the HFNC group. Survival analysis showed that the total and 48-h intubation rates were not significantly different. The ensemble AdaBoost decision tree model (internal and external validation set AUROC 0.878, 0.726) had the best predictive accuracy performance. The model Shapley algorithm showed Sequential Organ Failure Assessment (SOFA), acute physiology scores (APSIII), the minimum and maximum lactate value as risk factors for early failure and age, the maximum PaCO 2 and PH value, Glasgow Coma Scale (GCS), the minimum PaO 2 /FiO 2 ratio, and PaO 2 value as protective factors. Conclusion NIV was associated with lower intubation rate and ICU 28-day and in-hospital mortality. Further survival analysis reinforced that the effect of NIV on the intubation rate might partly be attributed to the other impact factors. The ensemble AdaBoost decision tree model may assist clinicians in making clinical decisions, and early organ function support to improve patients’ SOFA, APSIII, GCS, PaCO 2 , PaO 2 , PH, PaO 2 /FiO 2 ratio, and lactate values can reduce the early failure rate and improve patient prognosis.
Application of Machine Learning Algorithms to Identify Problematic Nuclear Data
In this work we aim to show that Machine learning algorithms are promising tools for the identification of nuclear data that contribute to increased errors in transport simulations. We demonstrate this through an application of a machine learning algorithm (Random Forest) to the Whisper/MCNP6 criticality validation library to identify nuclear data that are associated with an increase of the bias (simulated - experimental $k_{eff}$) in the calculations. Specifically, the $k_{eff}$ sensitivity profiles (w.r.t. nuclear data) of 233 U solution benchmarks are used to predict the bias and Shapley Additive Explanations (SHAP) are used to explain how the sensitivities are related to the predicted bias. The SHAP values can be interpreted as sensitivity coefficients of the machine learning model to the $k_{eff}$ sensitivities which are used to make predictions of bias. Using the SHAP values we can identify specific subsets of nuclear data which have the highest probability of influencing bias. We demonstrate the utility of this method by showing how SHAP values were used to identify an inconsistency in the 19 F inelastic scattering nuclear data. The methodology presented here is not limited to transport problems and can be applied to other simulations if there are experimental measurements to compare against, simulations of those experimental measurements, and the ability to calculate sensitivities of the model output with respect to the data inputs.
Accurate Machine Learning for Predicting the Viscosities of Deep Eutectic Solvents
Deep eutectic solvents (DESs) are emerging as environmentally friendly designer solvents for mass transport and heat transfer processes in industrial applications; however, the lack of accurate tools to predict and thus control their viscosities under both a range of environmental factors and formulations hinders their general application. While DESs may serve as designer solvents, with nearly unlimited combinations, this unfortunately makes it experimentally infeasible to comprehensively measure the viscosities of all DESs of potential industrial interest. To assist in the design of DESs, we have developed several new machine learning (ML) models that accurately and rapidly predict the viscosities of a diverse group of DESs at different temperatures and molar ratios using, to date, one of the most comprehensive data sets containing the properties of over 670 DESs over a wide range of temperatures (278.15–385.25 K). Three ML models, including support vector regression (SVR), feed forward neural networks (FFNNs), and categorical boosting (CatBoost), were developed to predict DES viscosity as a function of temperature and molar ratio and contrasted with multilinear and two-factor polynomial regression baselines. Further, quantum chemistry-based, COSMO-RS-derived sigma profile (σ-profile) features were used as inputs for the ML models. The CatBoost model is excellent at externally predicting DES viscosity, as indicated by high R 2 (0.99) and low root-mean-square-error (RMSE) and average absolute relative deviations (AARD) (5.22%) values for the testing data sets, and 98% of the data points lie within the 15% of AARD deviations. Furthermore, SHapley additive explanation (SHAP) analysis was employed to interpret the ML results and rationalize the viscosity predictions. The result is an ML approach that accurately predicts viscosity and will aid in accelerating the design of appropriate DESs for industrial applications.
Identifying Nuclear Data Correlated Through Predicting Bias in Integral Experiments via Applying Principal Component Analysis to Random Forest
ABSTRACT Nuclear data (ND) are the input data for neutron‐transport simulations to answer questions related to nuclear technologies. Subsets of ND, here > 20,000 data points, are validated with respect to thousands of criticality experiments that represent various applications on a small scale. The aim of validation with these experiments is to find errors in ND or methods. The key challenge here is that several hundreds of ND are used to simulate one integral value. Hence, one cannot clearly identify what ND are leading to bias in criticality measurements. In fact, a mistake in one nuclear‐data observable can be compensated with an error in another, and the predicted criticality value would still be predicted in agreement with experimental data. Random forest (RF) was previously employed to predict bias in criticality measurements using sensitivities of simulated criticality experiments to ND. The SHapley Additive exPlanations (SHAP) metric was then applied to attribute the importance of each ND experiment and observable to bias prediction. This, however, did not highlight what ND were jointly related to predicting bias. This is important as it could inform us about where compensating errors in ND could hide. We tackle this shortcoming here by first decomposing the ND sensitivities to integral‐experiment simulations into principal components. Then we use principal component projections to predict bias via the RF and SHAP. The SHAP values and principal components are employed to reconstruct detailed SHAP values for each ND observable. We demonstrate that these extended SHAP bias predictions are more robust, less noisy, and more efficient. In addition, we show that this approach accounts for covariance in ND sensitivities and automates the identification of where compensating errors could hide in ND.
High-Throughput Screening and Accurate Prediction of Ionic Liquid Viscosities Using Interpretable Machine Learning
Ionic liquids (ILs) are a novel group of green solvents with great promise for various industrial applications, including carbon capture and lignocellulosic biomass deconstruction. However, the use of ILs at the industrial scale remains challenging due to their high viscosities at ambient temperatures. To develop ILs with lower viscosities, a systematic study of their quantitative structure–property relationship (QSPR) is desirable. Here, we developed four machine learning (ML) models to predict viscosity at various temperature and pressure ranges, trained over a wide range of ILs consisting of various cationic and anionic families. ML methods including two-factor polynomial regression (two-factor PR), support vector regression (SVR), feed-forward neural networks (FFNN), and categorical boosting (CATBoost) were developed based on features that have proven useful in previous ML studies: COSMO-RS (conductor-like screening model for real solvents)-derived surface screening charge densities (sigma profiles). FFNN and CATBoost were the most accurate in predicting IL viscosities with lower average absolute relative deviation and higher R2 values on the test set. Tanimoto similarity scores were calculated to characterize the chemical space and structural similarity of the investigated ions. Furthermore, SHapley Additive exPlanation (SHAP) analysis was employed to interpret the ML results. Temperature, the polar area of ILs, and the nonpolar regions of ions are key features that influence the viscosity predictions. Importantly, the IL viscosity prediction here is the most accurate reported to date.
Reducing Southern Ocean Shortwave Radiation Errors in the ERA5 Reanalysis with Machine Learning and 25 Years of Surface Observations
Earth system models struggle to simulate clouds and their radiative effects over the Southern Ocean, partly due to a lack of measurements and targeted cloud microphysics knowledge. We have evaluated biases of downwelling shortwave radiation in the ERA5 climate reanalysis using 25 years (1995–2019) of summertime surface measurements, collected on the Research and Supply Vessel (RSV) Aurora Australis, the Research Vessel (R/V) Investigator, and at Macquarie Island. During October–March daylight hours, the ERA5 simulation of SW down exhibited large errors (mean bias = 54 W m -2 , mean absolute error = 82 W m -2 , root-mean-square error = 132 W m -2 , and R 2 = 0.71). To determine whether we could improve these statistics, we bypassed ERA5’s radiative transfer model for SW down with machine learning–based models using a number of ERA5’s gridscale meteorological variables as predictors. These models were trained and tested with the surface measurements of SW down using a 10-fold shuffle split. An extreme gradient boosting (XGBoost) and a random forest–based model setup had the best performance relative to ERA5, both with a near complete reduction of the mean bias error, a decrease in the mean absolute error and root-mean-square error by 25% ± 3%, and an increase in the R 2 value of 5% ± 1% over the 10 splits. Large improvements occurred at higher latitudes and cyclone cold sectors, where ERA5 performed most poorly. We further interpret our methods using Shapley additive explanations. Our results indicate that data-driven techniques could have an important role in simulating surface radiation fluxes and in improving reanalysis products.
Accelerating the design of lattice structures using machine learning
Lattices remain an attractive class of structures due to their design versatility; however, rapidly designing lattice structures with tailored or optimal mechanical properties remains a significant challenge. With each added design variable, the design space quickly becomes intractable. To address this challenge, research efforts have sought to combine computational approaches with machine learning (ML)-based approaches to reduce the computational cost of the design process and accelerate mechanical design. While these efforts have made substantial progress, significant challenges remain in (1) building and interpreting the ML-based surrogate models and (2) iteratively and efficiently curating training datasets for optimization tasks. Here, we address the first challenge by combining ML-based surrogate modeling and Shapley additive explanation (SHAP) analysis to interpret the impact of each design variable. We find that our ML-based surrogate models achieve excellent prediction capabilities (R 2 > 0.95) and SHAP values aid in uncovering design variables influencing performance. We address the second challenge by utilizing active learning-based methods, such as Bayesian optimization, to explore the design space and report a 5 × reduction in simulations relative to grid-based search. Collectively, these results underscore the value of building intelligent design systems that leverage ML-based methods for uncovering key design variables and accelerating design.
Effects of spatial variability in vegetation phenology, climate, landcover, biodiversity, topography, and soil property on soil respiration across a coastal ecosystem
Coastal terrestrial-aquatic interfaces (TAIs) are crucial contributors to global biogeochemical cycles and carbon exchange. A systematic evaluation of the interaction between coastal catchment properties and carbon dioxide (CO2) emission by soil respiration is significant for assessing carbon dynamics and predicting the future trajectory of atmospheric CO2 concentrations in coastal TAIs. The soil CO2 efflux in these transition zones is however poorly understood due to the high spatiotemporal dynamics of TAIs, as various sub-ecosystems in this region are compressed and expanded by complex influences of tides, changes in river levels, climate, and land use. We focus on the Chesapeake Bay region to (i) investigate the spatial heterogeneity of the coastal ecosystem and identify spatial zones with similar environmental characteristics based on the spatial data layers, including vegetation index (kNDVI), climate, landcover, diversity, topography, soil property, and relative tidal elevation; (ii) understand the primary driving factors affecting soil respiration within sub-ecosystems of the coastal ecosystem. Specifically, we employed hierarchical clustering analysis to identify spatial regions with distinct environmental characteristics, followed by the determination of main driving factors using Random Forest regression and SHapley Additive exPlanations. Maximum and minimum temperature are the main drivers common to all sub-ecosystems, while each region also has additional unique major drivers that differentiate them from one another. Precipitation exerts an influence on vegetated lands, while soil pH value holds importance specifically in forested lands. In croplands characterized by high clay content and low sand content, the significant role is attributed to bulk density. Wetlands demonstrate the importance of both elevation and sand content, with clay content being more relevant in non-inundated wetlands than in inundated wetlands. The topographic wetness index significantly contributes to the mixed vegetation areas, including shrub, grass, pasture, and forest. Additionally, our research reveals that dense vegetation land covers and urban/developed areas exhibit distinct soil property drivers. Overall, there is no one-size-fits-all approach to modeling carbon fluxes in coastal TAIs, and our study highlights the importance of further research and monitoring practices to improve our understanding of carbon dynamics and promote the sustainable management of coastal TAIs.
Surrogate model evaluation and building energy benchmarking for commercial buildings
Building energy consumption benchmarking involves challenges associated with various energy patterns for different building types; heating, ventilating, and air-conditioning (HVAC) system types; and climates. Given significant variation in energy use patterns, accurate prediction of long-term energy use using surrogate models remains challenging. Multiple linear regression (MLR) is commonly used for building energy benchmarking because of its simple structure; however, it lacks accuracy compared to other black-box models. Although many studies have compared surrogate models and offer guidance on model selection based on metrics, they do not provide detailed analysis on improving the surrogate model accuracy. In this paper, we implement a surrogate model using polynomial ridge regression (i.e., MLR with interaction terms combined with ridge regularization) for small office and retail strip mall buildings across six HVAC system types and all climate zones, for electricity and natural gas in baseline and proposed scenarios. A simulation workflow is developed using OpenStudio TM /EnergyPlus TM to generate simulation data using measures over a wide range of efficiency inputs. Enhancements based on statistical insights are used for improving the model accuracy using filters, input transformations, and change points. Surrogate models achieved average coefficient of variation of the root mean squared error (CVRMSE) values of 2.17, 1.06, 2.05, and 3.26 for proposed electricity, proposed natural gas, baseline electricity, and baseline natural gas, respectively, with enhancements reducing CVRMSE by an average of 14.9% across all combinations. We provide model interpretation via Shapley additive explanations to determine which input variables most influence energy consumption and provide supportive arguments for enhancements.