Relative Importance Of Features Research Articles

Effects of cultural landscape service features in national forest parks on visitors’ sentiments: A nationwide social media-based analysis in China

National Forest Parks (NFPs) represent the highest level of the forest park system in China and have long been considered to cultivate positive experiences among visitors through their rich cultural services. However, there is limited knowledge of the relationship between the cultural dimension of landscape services (CLSs) in NFPs and visitors’ positive experiences. This study represents the first effort to investigate the effects of NFP CLSs on visitors’ expressed sentiments and aims to further identify the relative importance of various CLS features. A total of 267 NFPs across 31 provinces in China were selected as case studies, and approximately 300,000 visitors’ online comments from a widely used travel platform (CTrip) were obtained for sentiment analysis using natural language processing. A novel systematic framework for assessing CLSs was proposed, encompassing five principal features and 19 indicator subcategories. Ordinary least-squares and spatial regression models were used to reveal associations between NFP CLSs and expressed sentiments, whereas the XGBoost model and Shapley Additive exPlanations (SHAP) values were employed to identify the relative importance of NFP CLS features. Three major findings were observed: First, the feature scoring of CLSs in NFPs was unevenly distributed across the country, with scores along the eastern seaboard relatively higher than those inland in the northwest. Second, beneficial associations between the five principal CLS features and visitors’ expressed sentiments were suggested. Third, the influence of landscape aesthetics was particularly prominent in promoting visitors’ positive sentiments, followed by education, physical activities and health, recreation and tourism, and cultural heritage and spiritual practices. These findings provide valuable insights for health-oriented NFP management, policies, and planning.

Using machine learning techniques to predict the risk of osteoporosis based on nationwide chronic disease data

Osteoporosis is a major public health concern that significantly increases the risk of fractures. The aim of this study was to develop a Machine Learning based predictive model to screen individuals at high risk of osteoporosis based on chronic disease data, thus facilitating early detection and personalized management. A total of 10,000 complete patient records of primary healthcare data in the German Disease Analyzer database (IMS HEALTH) were included, of which 1293 diagnosed with osteoporosis and 8707 without the condition. The demographic characteristics and chronic disease data, including age, gender, lipid disorder, cancer, COPD, hypertension, heart failure, CHD, diabetes, chronic kidney disease, and stroke were collected from electronic health records. Ten different machine learning algorithms were employed to construct the predictive mode. The performance of the model was further validated and the relative importance of features in the model was analyzed. Out of the ten machine learning algorithms, the Stacker model based on Logistic Regression, AdaBoost Classifier, and Gradient Boosting Classifier demonstrated superior performance. The Stacker model demonstrated excellent performance through ten-fold cross-validation on the training set and ROC curve analysis on the test set. The confusion matrix, lift curve and calibration curves indicated that the Stacker model had optimal clinical utility. Further analysis on feature importance highlighted age, gender, lipid metabolism disorders, cancer, and COPD as the top five influential variables. In this study, a predictive model for osteoporosis based on chronic disease data was developed using machine learning. The model shows great potential in early detection and risk stratification of osteoporosis, ultimately facilitating personalized prevention and management strategies.

A perceptual approach to address complex water management issues in lowland permeable catchments

Water quality management is a pressing global concern, and an increasingly complex issue due to climate and land-use change, legacy pollution, and the persistent release of well-known and emerging contaminants from diffuse and point sources. The increasing availability of high-frequency monitoring data is leading existing, often heuristic approaches, to be judged inadequate. Water managers frequently rely on simple qualitative and/or quantitative approaches for decision-making, but a lack of tangible improvement in freshwater quality outcomes is demanding new and innovative approaches that rely more on physical process understanding, rather than precedent. In this study, we drew upon local geological, hydrogeological, and hydrological data to infer a high-level perceptual model of surface/groundwater interactions in a chalk stream in Dorset, UK. We used the perceptual model to interrogate spatial and temporal trends in historical water quality data and to construct reach-scale nutrient mass balances. Through novel representation with Sankey diagrams, the perceptual model highlighted the relative importance of different hydrological features. Surface/groundwater interactions were found to occur predominantly by spring flow. We demonstrate that river flow accretion was dominated by the Chalk aquifer despite only occupying ca. 15 % of the surface bedrock area, and that spring sources, whilst vital to dilute treated sewage inputs in baseflow conditions, were also major sources of legacy nitrate. Nutrient mass balances showed that sewage treatment works contributed ca. 13 % to soluble reactive phosphorus load, with groundwater accounting for ca. 48 %. Thus, a determinand often associated with point-source pollution was shown to be diffuse dominated in this river. The study demonstrated how a multi-disciplinary approach to water management, based on a comprehensive perceptual modelling approach, could identify hitherto unknown sources and relative contributors to freshwater pollution and allow flow and load apportionment studies to provide useful decision-support to manage nutrient pollution. The novel application of perceptual modelling tools, such as the Sankey diagram, allows different source attributions to be presented in an accessible manner, and can be readily transferred to other study areas.

Designing outdoor fitness areas for older adults: a conjoint analysis study

ABSTRACT This study aimed to examine the relative importance of micro-level design features of outdoor fitness areas preferred by older adults and whether preferences varied according to gender, park accompaniment, and mobility status. A series of Adaptive Choice-Based Conjoint (ACBC) tasks were completed by older adults (n = 372, 65–93 years, 57% female) in an online survey. Ten different micro-level design features of outdoor fitness areas were examined including surface type, location of fitness equipment, equipment designed for older adults, equipment distribution, shade sails, light fixtures, shady trees, benches, drinking water, and water body near fitness area. Hierarchical Bayes analyses within Sawtooth software were used to identify the relative importance scores and part-worth utilities of the design features. For the overall sample, the three most important features were as follows: surface type (21.5%, 95% CI 20.0, 22.9); equipment specifically designed for older adults (18.7%, 95% CI 17.3, 20.2); and shady trees around the fitness area (14.0%, 95% CI 12.9, 15.0). A few significant differences in the relative ranking of design features were observed by gender and mobility status. This study will help stakeholders in understanding which design features to prioritise when (re)designing outdoor fitness areas for older adults.

Machine Learning-Based Discrimination of Cardiovascular Outcomes in Patients With Hypertrophic Cardiomyopathy

BackgroundCurrent risk stratification strategies for patients with hypertrophic cardiomyopathy (HCM) are limited to traditional methodologies. ObjectivesThe authors aimed to establish machine learning (ML)-based models to discriminate major cardiovascular events in patients with HCM. MethodsWe enrolled consecutive HCM patients from 2 tertiary referral centers and used 25 clinical and echocardiographic features to discriminate major adverse cardiovascular events (MACE), including all-cause death, admission for heart failure (HF-adm), and stroke. The best model was selected for each outcome using the area under the receiver operating characteristic curve (AUROC) with 20-fold cross-validation. After testing in the external validation cohort, the relative importance of features in discriminating each outcome was determined using the SHapley Additive exPlanations (SHAP) method. ResultsIn total, 2,111 patients with HCM (age 61.4 ± 13.6 years; 67.6% men) were analyzed. During the median 4.0 years of follow-up, MACE occurred in 341 patients (16.2%). Among the 4 ML models, the logistic regression model achieved the best AUROC of 0.800 (95% CI: 0.760-0.841) for MACE, 0.789 (95% CI: 0.736-0.841) for all-cause death, 0.798 (95% CI: 0.736-0.860) for HF-adm, and 0.807 (95% CI: 0.754-0.859) for stroke. The discriminant ability of the logistic regression model remained excellent when applied to the external validation cohort for MACE (AUROC = 0.768), all-cause death (AUROC = 0.750), and HF-adm (AUROC = 0.806). The SHAP analysis identified left atrial diameter and hypertension as important variables for all outcomes of interest. ConclusionsThe proposed ML models incorporating various phenotypes from patients with HCM accurately discriminated adverse cardiovascular events and provided variables with high importance for each outcome.

Neuroimaging data repositories and AI-driven healthcare-Global aspirations vs. ethical considerations in machine learning models of neurological disease.

Neuroimaging data repositories are data-rich resources comprising brain imaging with clinical and biomarker data. The potential for such repositories to transform healthcare is tremendous, especially in their capacity to support machine learning (ML) and artificial intelligence (AI) tools. Current discussions about the generalizability of such tools in healthcare provoke concerns of risk of bias-ML models underperform in women and ethnic and racial minorities. The use of ML may exacerbate existing healthcare disparities or cause post-deployment harms. Do neuroimaging data repositories and their capacity to support ML/AI-driven clinical discoveries, have both the potential to accelerate innovative medicine and harden the gaps of social inequities in neuroscience-related healthcare? In this paper, we examined the ethical concerns of ML-driven modeling of global community neuroscience needs arising from the use of data amassed within neuroimaging data repositories. We explored this in two parts; firstly, in a theoretical experiment, we argued for a South East Asian-based repository to redress global imbalances. Within this context, we then considered the ethical framework toward the inclusion vs. exclusion of the migrant worker population, a group subject to healthcare inequities. Secondly, we created a model simulating the impact of global variations in the presentation of anosmia risks in COVID-19 toward altering brain structural findings; we then performed a mini AI ethics experiment. In this experiment, we interrogated an actual pilot dataset (n = 17; 8 non-anosmic (47%) vs. 9 anosmic (53%) using an ML clustering model. To create the COVID-19 simulation model, we bootstrapped to resample and amplify the dataset. This resulted in three hypothetical datasets: (i) matched (n = 68; 47% anosmic), (ii) predominant non-anosmic (n = 66; 73% disproportionate), and (iii) predominant anosmic (n = 66; 76% disproportionate). We found that the differing proportions of the same cohorts represented in each hypothetical dataset altered not only the relative importance of key features distinguishing between them but even the presence or absence of such features. The main objective of our mini experiment was to understand if ML/AI methodologies could be utilized toward modelling disproportionate datasets, in a manner we term "AI ethics." Further work is required to expand the approach proposed here into a reproducible strategy.

Improving Academic Advising in Engineering Education with Machine Learning Using a Real-World Dataset

The role of academic advising has been conducted by faculty-student advisors, who often have many students to advise quickly, making the process ineffective. The selection of the incorrect qualification increases the risk of dropping out, changing qualifications, or not finishing the qualification enrolled in the minimum time. This study harnesses a real-world dataset comprising student records across four engineering disciplines from the 2016 and 2017 academic years at a public South African university. The study examines the relative importance of features in models for predicting student performance and determining whether students are better suited for extended or mainstream programmes. The study employs a three-step methodology, encompassing data pre-processing, feature importance selection, and model training with evaluation, to predict student performance by addressing issues such as dataset imbalance, biases, and ethical considerations. By relying exclusively on high school performance data, predictions are based solely on students’ abilities, fostering fairness and minimising biases in predictive tasks. The results show that removing demographic features like ethnicity or nationality reduces bias. The study’s findings also highlight the significance of the following features: mathematics, physical sciences, and admission point scores when predicting student performance. The models are evaluated, demonstrating their ability to provide accurate predictions. The study’s results highlight varying performance among models and their key contributions, underscoring the potential to transform academic advising and enhance student decision-making. These models can be incorporated into the academic advising recommender system, thereby improving the quality of academic guidance.

Classification of cotton water stress using convolutional neural networks and UAV-based RGB imagery

<p>Embracing smart irrigation management techniques empowers growers to irrigate with greater efficiency, thereby promoting sustainable agricultural production. In this context, growers often rely on crop evapotranspiration (ETc) as a key factor in making informed irrigation decisions, underscoring the significance of accurately determining and spatially mapping crop water status. Technological progress, exemplified by the emergence of unmanned aerial vehicles (UAVs), has brought about a revolutionary shift in agricultural monitoring. UAV platforms can capture high-resolution images with centimeter-level spatial accuracy and offer higher temporal coverage compared to satellite imagery. Considering these advancements, this study introduces a robust method for classifying water stress in cotton using a compact UAV platform and convolutional neural networks (CNN). The experiment was conducted at the USDA-ARS Cropping Systems Research Laboratory (CSRL) in Lubbock, Texas, where the cotton field was divided into 12 drip zones. The study included three replications to evaluate four irrigation treatments: “rainfed”, “full irrigation”, “percent deficit of full irrigation”, and “time delay of full irrigation”. The results demonstrated that the CNN model successfully classified the cotton water stress using the UAV-based RGB image, achieving an overall best prediction accuracy of approximately 91%. By segmenting the original cotton images into separate canopy and soil areas using morphological image processing methods, the authors also isolated and analyzed the individual contributions of these components to cotton water stress. Additionally, a random forest classifier revealed the relative importance of different image features in the classification process through feature importance analysis. These findings highlighted the state-of-the-art performance of the proposed system in cotton water stress classification and provided valuable insights into the key image features contributing to accurate classification. The authors concluded that integrating UAV-based RGB imagery and CNN models had great potential for assessing water stress in cotton.<strong></strong></p>

Classification of cotton water stress using convolutional neural networks and UAV-based RGB imagery

<p>Embracing smart irrigation management techniques empowers growers to irrigate with greater efficiency, thereby promoting sustainable agricultural production. In this context, growers often rely on crop evapotranspiration (ETc) as a key factor in making informed irrigation decisions, underscoring the significance of accurately determining and spatially mapping crop water status. Technological progress, exemplified by the emergence of unmanned aerial vehicles (UAVs), has brought about a revolutionary shift in agricultural monitoring. UAV platforms can capture high-resolution images with centimeter-level spatial accuracy and offer higher temporal coverage compared to satellite imagery. Considering these advancements, this study introduces a robust method for classifying water stress in cotton using a compact UAV platform and convolutional neural networks (CNN). The experiment was conducted at the USDA-ARS Cropping Systems Research Laboratory (CSRL) in Lubbock, Texas, where the cotton field was divided into 12 drip zones. The study included three replications to evaluate four irrigation treatments: “rainfed”, “full irrigation”, “percent deficit of full irrigation”, and “time delay of full irrigation”. The results demonstrated that the CNN model successfully classified the cotton water stress using the UAV-based RGB image, achieving an overall best prediction accuracy of approximately 91%. By segmenting the original cotton images into separate canopy and soil areas using morphological image processing methods, the authors also isolated and analyzed the individual contributions of these components to cotton water stress. Additionally, a random forest classifier revealed the relative importance of different image features in the classification process through feature importance analysis. These findings highlighted the state-of-the-art performance of the proposed system in cotton water stress classification and provided valuable insights into the key image features contributing to accurate classification. The authors concluded that integrating UAV-based RGB imagery and CNN models had great potential for assessing water stress in cotton.<strong></strong></p>

Stellar Atmospheric Parameters for Cool Dwarfs in Gaia Data Release 3

We provide a catalog of atmospheric parameters for 1,806,921 cool dwarfs from Gaia Data Release 3 (DR3) that lie within the range covered by LAMOST cool dwarf spectroscopic parameters: 3200 K < T eff < 4300 K, −0.8 < [M/H] < 0.2 dex, and 4.5 < log g < 5.5 dex. Our values are derived based on machine-learning models trained with multiband photometry corrected for dust. The photometric data comprise optical data from the Sloan Digital Sky Survey r, i, and z bands, near-infrared data from the Two Micron All Sky Survey J, H, and K bands, and mid-infrared data from the ALLWISE W1 and W2 bands. We used both random forest and light gradient boosting machine machine-learning models and found similar results from both, with an error dispersion of 68 K, 0.22 dex, and 0.05 dex for T eff, [M/H], and log g, respectively. Assessment of the relative feature importance of different photometric colors indicated W1 − W2 as most sensitive to both T eff and log g, with J − H being most sensitive to [M/H]. We find that our values show a good agreement with the Apache Point Observatory Galactic Evolution Experiment, but are significantly different to those provided as part of Gaia DR3.

Elucidating the response mechanisms of hydrological and sediment connectivity to the river network structure, vegetation, and topographic features in the Jinghe River Basin

AbstractThe management of basin water and soil resources greatly benefits from investigating the spatial changes and primary determinants of hydrological and sediment connectivity from perspectives of topographic features, vegetation characteristics, structural features of river networks. However, quantifying the effects of influencing factors and their interactions on connectivity is still a challenge in the field of studies of surface processes. To address this challenge, we applied the geographical detector model (GDM) and random forest (RF) to quantify the relative importance and explanatory power of topographic factors, vegetation features, and river network structure on the hydrological connectivity and sediment connectivity and to clarify the interaction mechanism of various factors in the Jinghe River Basin. Results indicate that from 2005 to 2020, the mean value of hydrological connectivity (ICH) witnessed a decreasing trend from −5.47 to −5.58. Similarly, the mean value of sediment connectivity (ICS) declined over the same period from −9.74 to −10.05, with high values registered in valleys and low values in plains. The river density (D) and hydrological weight factor (normalized SCS‐CN) exhibited a greater spatial explanatory power on ICH than other parameters, reaching 0.149 and 0.182, respectively. The vegetation factor (0.299) and sediment weight factor (0.410) manifested considerably more influence over ICS. Spatial integration between relevant elements can enhance our understanding of basin connectivity. Interactions between vegetation, topography, and river network structure provide more explanatory power compared to interactions among same‐type features. Ultimately, this study offers a theoretical template for understanding the management of soil and water resources in the Jinghe River Basin, as well as the spatial variability in soil erosion.

Investigating the applicability of deep learning and machine learning models in predicting the structural performance of V-shaped RC folded plates

Reinforced concrete folded plates (RC-FPs), which are a special class of shell structures, have recently become very popular in modern architectural and engineering applications because of the need for lightweight and aesthetic structures to cover large areas. However, it is known that studies on the structural performance of RC-FPs are insufficient. Therefore, this article presents a study on the development and comparison of different deep learning (DL) and machine learning (ML) models for the prediction of the structural performance of full-scale V-shaped RC-FPs produced from hybrid fiber-reinforced self-consolidating concrete (HFR-SCC) having different plate thicknesses (50, 60, 70, and 80 mm), fiber volumes (1.00% and 1.25%), and combinations (single, binary, and ternary). While vanilla long short-term memory (VLSTM) and bidirectional long short-term memory (BILSTM) are used as DL models, random forest (RF), extremely randomized trees (ERT), and adaptive boosting (AdaBoost) are preferred for ML models. To construct the models, the structural performance results of a total of 44 full-scale V-shaped RC-FPs subjected to four-point bending loading were adopted as the database. In addition to all these, the AdaBoost model is used to determine the relative feature importance of the input parameters. Based on the results, among the DL models, the BILSTM has the best ability to predict the structural performance values of V-shaped RC-FPs (such as R-squared values for maximum load-bearing capacity, cracking load, toughness, and deflection ductility are 0.934, 0.987, 0.972, and 0.812, respectively), while in ML models, this is valid for the ERT (such as R-square values are 0.917 for maximum load-bearing capacity, 0.936 for cracking load, 0.947 for toughness and 0.825 for deflection ductility). On the other hand, DL models predicted all other structural performance values better than ML models, except for deflection ductility. Besides, the most relative important input features for maximum load-bearing capacity and toughness values is plate thickness, whereas for cracking load and deflection ductility values compressive strength is important. In conclusion, it can be emphasized that the use of DL models can provide significant advantages in engineering applications, such as predicting the structural performance of V-shaped RC-FPs.

Predicting the bandgap and efficiency of perovskite solar cells using machine learning methods.

Rapid and accurate prediction of bandgaps and efficiency of perovskite solar cells is a crucial challenge for various solar cell applications. Existing theoretical and experimental methods often accurately measure these parameters; however, these methods are costly and time-consuming. Machine learning-based approaches offer a promising and computationally efficient method to address this problem. In this study, we trained different machine learning(ML) models using previously reported experimental data. Among the different ML models, the CatBoostRegressor performed better for both bandgap and efficiency approximations. We evaluated the proposed model using k-fold cross-validation and investigated the relative importance of input features using Shapley Additive Explanations (SHAP). SHAP interprets valuable insights into feature contributions of the prediction of the proposed model. Furthermore, we validated the performance of the proposed model using an independent dataset, demonstrating its robustness and generalizability beyond the training data. Our findings show that machine learning-based approaches, with the aid of SHAP, can provide a promising and computationally efficient method for the accurate and rapid prediction of perovskite solar cell properties. The proposed model is expected to facilitate the discovery of new perovskite materials and is freely available at GitHub (https://github.com/AsadKhanJBNU/perovskite_bandgap_and_efficiency.git) for the perovskite community.

Application of machine learning techniques to profile smoking behavior of adolescent girls in Ghana

Background Tobacco use trends among adolescents in low- and middle-income countries, and in particular narrowing gender gaps, highlight the need for interventions to prevent and/or reduce tobacco use among adolescent girls. We evaluated a social marketing program in Ghana discouraging tobacco use among adolescent girls and additionally investigated the pathways influencing smoking behaviors to identify programmatic opportunities for impact. Leveraging the data collected through the stepped wedge cluster randomized trial and panel survey of 9000 girls aged 13–19 , we sought to apply machine learning (ML) techniques to identify the most important variables for predicting initiation of smoking. Methods To identify predictors of smoking initiation we sought to develop a model which could accurately differentiate smokers from non-smokers and evaluated various ML approaches for training classifier algorithms to achieve this. We selected a Synthetic Minority Over-sampling Technique (SMOTE) because it optimized the recall and precision of the model. We then utilized the technique of feature importance for greater insight into how the model arrived at its decisions and to rank the most important variables for predicting smokers. To explore different dimensions of smoking behavior, including initiation and continuation, we trained our model by using several combinations of target outcomes and input variables from the panel survey. Results The resulting features of smokers highlight the importance of girls’ independence and connectivity, social environment, and peer influence on likelihood of smoking, and in particular subsequent initiation. These results were largely consistent with our formative research findings based on qualitative interviews informed by behavioral science. Conclusions This novel application of ML techniques demonstrates how data science approaches can generate new programmatic insights from rigorous evaluation data, especially when data collection is informed by behavioral theory. Such insights about the relative importance of different features can be valuable input for program planning and outreach.

Relative importance of vegetation features and intra- and inter-specific interactions on habitat preferences of a raptor guild in eucalypt plantations

Forest plantations are increasingly recognized for their role in providing ecosystem services in agroforestry systems. Understanding the habitat preferences of potentially beneficial birds supports effective forest management. Our study focused on identifying breeding habitat preferences of a raptor guild and assessing exotic eucalypt plantations as potential breeding sites. We explored preferences at various spatial scales considering vegetation features, intraspecific territorial behaviour, and interspecific interactions within the guild to determine their importance. We hypothesized that feeding habits and body mass explain differences in habitat preferences among three diurnal raptors (Northern goshawk, Common buzzard, Eurasian sparrowhawk). We employed random forest models to differentiate breeding sites from control plots, assessing factors' relative importance and unravelling the hierarchical process of habitat selection in southwestern Europe's eucalypt plantations. The studied eucalypt plantations emerged as suitable breeding habitats, harbouring dense breeding populations with uniform spatial distribution. Intraspecific territoriality and vegetation significantly influenced breeding habitat selection, while interspecific interactions held lesser influence. Species-specific variations were attributed to feeding habits and body masses. Tall eucalypt trees and mature-like patches were crucial elements driving breeding habitat selection. Raptors' preference for specific woody elements primarily influences areas within a relatively small radius around their nests (including the nest tree, nest site −10 m radius, and nest stand −50 m), with sparrowhawks showing preferences extending to 125 m. Raptors' inclination for mature woody elements and mature forest patches should steer forest management to bolster these apex predators and their contributions in eucalypt plantations within agroforestry systems.

On the failings of Shapley values for explainability

Explainable Artificial Intelligence (XAI) is widely considered to be critical for building trust into the deployment of systems that integrate the use of machine learning (ML) models. For more than two decades Shapley values have been used as the theoretical underpinning for some methods of XAI, being commonly referred to as SHAP scores. Some of these methods of XAI now rank among the most widely used, including in high-risk domains. This paper proves that the existing definitions of SHAP scores will necessarily yield misleading information about the relative importance of features for predictions. The paper identifies a number of ways in which misleading information can be conveyed to human decision makers, and proves that there exist classifiers which will yield such misleading information. Furthermore, the paper offers empirical evidence that such theoretical limitations of SHAP scores are routinely observed in ML classifiers.

Real-Time LNG Buses Emissions Prediction Based on a Temporal Fusion Trans-Formers Model

Emissions from transportation are one of the key factors preventing the achievement of carbon peaking and carbon neutrality by 2050, with particular attention to emissions from buses. Specifically, few research has been conducted on the exhaust emissions characteristics of liquified natural gas (LNG) buses under different driving scenarios. This study proposed a framework for predicting exhaust emissions of LNG buses based on the portable emission measurement system and GPS collaborative perception data. Firstly, the emission distribution characteristics of CO2, CO, HC, and NOx from LNG buses in real-world driving were analyzed by visualization methods. Then, the real-time exhaust emissions of LNG buses were predicted based on the temporal fusion transformers model for both urban and suburban sections of Zhenjiang City, and the model validity was verified. The current and past 10 s driving states were used for predicting the emission rate of LNG buses. The results showed that the proposed model outperforms other advanced algorithms in real-time exhaust emissions prediction of LNG buses, with an average R2 value higher than 0.94 and an average MAPE reduction of 14.19%. The error assessment revealed that the emission values and average emission rates are higher when driving in the urban section compared to the suburban section. Among the influencing factors, traffic conditions have the most significant impacts on the exhaust emissions of LNG buses, followed by road conditions and driving states, with relative feature importance of 48.9, 34.8, and 16.3%, respectively. Additionally, the current and past 10 s driving states also significantly influenced real-time predictions. This study provides an essential theoretical reference for reducing exhaust emissions for city buses.

Data driven approach for weight restricted data envelopment analysis models with single output

This study aims to explore whether a machine learning algorithm can be used to make improvements in assessing unit efficiencies via a data envelopment analysis (DEA) model. In this study, a DEA model is used to calculate the efficiency scores of Desicion Making Units (DMUs). Then, an ML algorithm is trained that aims to predict the single output using inputs. Ranking of input features based on relative feature importance values obtained from the trained ML model is fed to the DEA model as weight restrictions. As a result, the two DEA models are compared with each other. ML-based insights (feature importance ranking) improve the DEA model in the direction of fewer zero weights. The additional weight restrictions are data depdendent, and hence realistic. As a novel approach, this study proposes the use of machine learning-based feature importance values to overcome a limitation of a DEA model.

A statistical perspective for predicting the strength of metals: Revisiting the Hall–Petch relationship using machine learning

The mechanical properties of a material are intimately related to its microstructure. This is particularly important for predicting mechanical behavior of polycrystalline metals, where microstructural variations dictate the expected material strength. Until now, the lack of microstructural variability in available datasets precluded the development of robust physics-based theoretical models that account for randomness of microstructures. To address this, we have developed a probabilistic machine learning framework to predict the flow stress as a function of variations in the microstructural features. In this framework, we first generated an extensive database of flow stress for a set of over a million randomly sampled microstructural features, and then applied a combination of mixture models and neural networks on the generated database to quantify the flow stress distribution and the relative importance of microstructural features. The results show excellent agreement with experiments and demonstrate that across a wide range of grain size, the conventional Hall–Petch relationship is statistically valid for correlating the strength to the average grain size and its comparative importance versus other microstructural features. This work demonstrates the power of the machine-learning based probabilistic approach for predicting polycrystalline strength, directly accounting for microstructural variations, resulting in a tool to guide the design of polycrystalline metallic materials with superior strength, and a method for overcoming sparse data limitations.

Experimental results on data analysis algorithms for extracting and interpreting edge feature data for duct tape and textile physical fit examinations.

A physical fit is an important observation that can result from the forensic analysis of trace evidence as it conveys a high degree of association between two items. However, physical fit examinations can be time-consuming, and potential bias from analysts may affect judgment. To overcome these shortcomings, a data analysis algorithm using mutual information and a decision tree has been developed to support practitioners in interpreting the evidence. We created these tools using data obtained from physical fit examinations of duct tape and textiles analyzed in previous studies, along with the reasoning behind the analysts' decisions. The relative feature importance is described by material type, enhancing the knowledge base in this field. Compared with the human analysis, the algorithms provided accuracies above 90%, with an improved rate of true positives for most duct tape subsets. Conversely, false positives were observed in high-quality scissor cut (HQ-HT-S) duct tape and textiles. As such, it is advised to use these algorithms in tandem with human analysis. Furthermore, the study evaluated the accuracy of physical fits when only partial sample lengths are available. The results of this investigation indicated that acceptable accuracies for correctly identifying true fits and non-fits occurred when at least 35% of a sample length was present. However, lower accuracies were observed for samples prone to stretching or distortion. Therefore, the models described here can provide a valuable supplementary tool but should not be the sole means of evaluating samples.