Search

Distraction is one of the most prominent driver characteristics which accounts for several road crashes and is a major cause of concern for almost all road safety institutions. The main objective of this study was to explore and identify various causes of a crash on Yamuna Expressway. Furthermore, another objective includes the detailed study of driver's fault, including in-vehicular distraction and identifying the most and least distracting activity. Moreover, other driver characteristics like fatigue, the reaction of the driver when someone suddenly comes on the way, etc. were also analyzed for their contributions towards crash occurrence. In this study, supervised machine learning techniques viz. Classification and Regression Tree (CART) was used to predict the possibility of crash occurrence. The model results show that using a cell phone in case of adverse weather conditions has the highest likelihood of crash occurrence. Moreover, among other distraction tasks adjustment of radio/compact disc (CD), while driving has the least likelihood of a crash. Further, it was also found that the reaction of the driver when someone comes suddenly on the way (e.g. animal, another crossing vehicle/pedestrian) and long driving hours leading to headache/backache/fatigue also affect the possibility of crash occurrence. Model results were then validated using the testing dataset and it was found that the model is equally efficient in predicting crashes as it was with the training dataset. This study provides insight into several important driver characteristics and will be helpful in developing educational and enforcement strategies thereby reducing crash risk. [ABSTRACT FROM AUTHOR]

The main clinical intervention for older adults with hearing loss is the provision of hearing aids. However, uptake and usage in this population have historically been reported as low. The aim of this study was to understand the hearing loss characteristics, from measured audiometric hearing loss and self-perceived hearing handicap, that contribute to the decision of hearing aid ownership. A total of 2833 adults aged 50+ years, of which 329 reported hearing aid ownership, were involved with a population-based survey with audiometric hearing assessments. Classification and regression tree (CART) analysis was used to classify hearing aid ownership from audiometric measurements and hearing disability outcomes. An overall accuracy of 92.5% was found for the performance of the CART analysis in predicting hearing aid ownership from hearing loss characteristics. By including hearing disability, sensitivity for predicting hearing aid ownership increased by up to 40% compared with just audiometric hearing loss measurements alone. A decision tree approach that considers both objectively measured hearing loss and self-perceived hearing disability, could facilitate a more tailored and personalised approach for determining hearing aid needs in the older population. Without intervention, older adults with hearing loss are at higher risk of cognitive decline and higher rates of depression, anxiety, social isolation. The provision of hearing aids can compensate hearing function, however, uptake and usage have been reported as low. Using a more precise cut-off from audiometric measures and self-perceived hearing disability scores could facilitate a tailored and personalised approach to screen and identify older adults for hearing aid needs. [ABSTRACT FROM AUTHOR]

In the United States, the probability of work zone crashes has increased due to an increase in renovation works by transportation infrastructures. The severity of work zone crashes is associated with multiple contributing factors such as the roadway's geometric design features, temporal variables, environmental conditions, types of vehicles, and driver behaviors. For this study, we acquired and analyzed three years (2016–2018) of work zone crash data from the state of New Jersey. We investigated the performance of several machine learning methods, including Support Vector Machine, Random Forest, Catboost, Light GBM, and XGBoost to predict the type of injury severity resulting from work zone crashes. To evaluate models' performances, some statistical evaluation parameters such as accuracy, precision, and recall scores were calculated. In addition, a sensitivity analysis was conducted to assess the impact of the most influential factors in work zone-related crashes. Random Forest and Catboost outperformed the other models in terms of predicting fatal, major, and minor injuries. According to the sensitivity analysis, crash type and speed limit were the most significantly associated variables with crash severity. The findings of this study are expected to facilitate the identification of appropriate countermeasures for reducing the severity of work zone crashes. [ABSTRACT FROM AUTHOR]

The present study aimed to compare five machine learning techniques, namely, artificial neural network (ANN), support vector machine (SVM), chi-square automatic interaction detection (CHAID), classification and regression tree (CART), and random forest (RF) for mass appraisal of real estate. Firstly, 1982 precedent data was collected throughout the entire study area for train and test models. Secondly, a total of 68 variables were considered for the mass appraisal. Subsequently, the five machine learning techniques were applied. Finally, the receiver operating characteristic (ROC) and various statistical methods were applied to compare five machine learning techniques. [ABSTRACT FROM AUTHOR]

Despite the high level of interpretability of the Classification and Regression Tree (CART) as one of the widely-used data mining models, its results usually are less accurate than other models. To address the low accuracy of the CART model, this paper developed two new and enhanced CART-based algorithms that can provide both high interpretability and high accuracy for flood susceptibility mapping. In the developed models, various hyperparameters of the CART were explicitly tuned using the genetic algorithm (GA) and grid search (GS) methods. Eleven variables and 222 flood locations were used in Sardabroud watershed in Iran for modelling. The results showed area under the receiver operating characteristic curve (AUROC) values of 0.884, 0.927 and 0.908, respectively, for the CART, CART-GS and CART-GA models in the validation run. The results showed that the developed models can produce the output that satisfies both accuracy and interpretability criteria. [ABSTRACT FROM AUTHOR]

Background and Objective: Confounding factors should be considered when predicting visual prognosis after epiretinal membrane (ERM) surgery. We aimed to predict visual acuity after ERM surgery using regression tree analysis. Patients and Methods: We retrospectively reviewed 343 eyes of 343 patients who underwent vitrectomy for idiopathic ERM. Regression tree analysis was performed to predict best-corrected visual acuity (BCVA) at 6 months postoperatively. Results: Patients were first divided into two groups based on preoperative BCVA. Age, central macular thickness, and axial length were associated with visual prognosis in patients with specific preoperative BCVAs. When patients were subdivided into groups 1–8, BCVA was significantly better in group 1 than in groups 4–8 (p≤0.001), and in groups 2–5 and 7 than in groups 6 and 8 (p≤0.002). Conclusion: Surgical prognosis in eyes with ERM should be evaluated based on a comprehensive assessment of preoperative characteristics. [ABSTRACT FROM AUTHOR]

Scour around piles is considered as one of the most important phenomena which could undermine the stability of piles. Data-driven methods are now being used more and more to predict complex phenomena such as scour around of pile groups. In this study, three data-driven methods including multivariate adaptive regression spline (MARS), classification and regression tree (CART) and artificial neural network (ANN), are used for determining current-induced scour depth around pile groups in clear water conditions. Obtained results are compared with those of M5 Model Tree (M5MT) which shows the higher precision of them. The calculated statistical indices revealed the higher accuracy of MARS in prediction of scour depth (correlation coefficient [CC] = 0.872 and root mean square error [RMSE] = 0.233) in comparison with ANN (CC = 0.843 and RMSE = 0.256), CART (CC = 0.734 and RMSE = 0.323) and M5MT (CC = 0.761 and RMSE = 0.404) for testing data. Furthermore, the developed discrepancy ratio (DDR) index confirms the outperformance of MARS in comparison with other approaches for prediction of scour depth. Finally, the parametric investigations conducted through MARS show that the parameter of G / D has the highest effect on the scour depth. This finding is in agreement with the results of CART. [ABSTRACT FROM AUTHOR]

Clustered binary outcomes are frequently encountered in clinical research (e.g. longitudinal studies). Generalized linear mixed models (GLMMs) for clustered endpoints have challenges for some scenarios (e.g. data with multi-way interactions and nonlinear predictors unknown a priori). We develop an alternative, data-driven method called Binary Mixed Model (BiMM) tree, which combines decision tree and GLMM within a unified framework. Simulation studies show that BiMM tree achieves slightly higher or similar accuracy compared to standard methods. The method is applied to a real dataset from the Acute Liver Failure Study Group. [ABSTRACT FROM AUTHOR]

Departments of Transportation regularly evaluate the condition of pavements through visual inspections, nondestructive evaluations, image recognition models and learning algorithms. The above methodologies, though efficient, have drawn attention due to their subjective errors, uncertainties, noise effects and overfitting. To improve on the outcomes of the shallow learning models already used in pavement crack prediction, this paper reports on an investigation of the use of recursive partitioning and artificial neural networks (ANN; deep learning frameworks) in predicting the crack rating of pavements. Explanatory variables such as the average daily traffic and truck factor, roadway functional class, asphalt thickness, and pavement condition time series data are employed in the model formulation. Overall, it is observed that the recursive partitioning (regression tree – R2 > 0.8 and classification tree – R2 > 0.6) and ANN (continuous response – R2 > 0.8 and categorical response – R2 > 0.6) are compelling machine learning models for the prediction of the crack ratings based on their goodness-of-fit statistics, mean absolute deviation (MAD < 0.4) and the root mean square errors (RMSE between 0.30 and 0.65). [ABSTRACT FROM AUTHOR]

Accurate knowledge of equilibrium conditions for methane hydrates dissociation is crucial to implement an appropriate hydrate-based technology/process. Hence, much importance is attached to utilize techniques for the modeling of hydrate stability zones and the existence of reliable data as well. This research employs the published experimental data, from 1940 to 2016, for modeling the incipient stability conditions of methane hydrate in pure water using Classification and Regression Tree (CART) as a novel methodology in gas hydrates research. Least squares support vector machine (LSSVM) and artificial neural network (ANN) methods were selected as a basis for comparison. Furthermore, the Leverage mathematical approach was used for evaluating the quality of the data as well as the correctness of the CART model. Results confirmed that the developed tree-based model provides excellent outcomes and no model can rival it for accuracy. In addition, applying the Leverage algorithm specified that: (a) the CART model is statistically valid and correct; (b) the published experimental data in the literature are different in quality; (c) there is no doubtful data in the studied database; and (d) the published data in the temperature range from 148.8 to 238.8 K are not sufficient compared with the whole of the collected database. [ABSTRACT FROM AUTHOR]

Objective: Guaranteeing a safe and comfortable driving workload can contribute to reducing traffic injuries. In order to provide safe and comfortable threshold values, this study attempted to classify driving workload from the aspects of human factors mainly affected by highway geometric conditions and to determine the thresholds of different workload classifications. This article stated a hypothesis that the values of driver workload change within a certain range.Methods: Driving workload scales were stated based on a comprehensive literature review. Through comparative analysis of different psychophysiological measures, heart rate variability (HRV) was chosen as the representative measure for quantifying driving workload by field experiments. Seventy-two participants (36 car drivers and 36 large truck drivers) and 6 highways with different geometric designs were selected to conduct field experiments. A wearable wireless dynamic multiparameter physiological detector (KF-2) was employed to detect physiological data that were simultaneously correlated to the speed changes recorded by a Global Positioning System (GPS) (testing time, driving speeds, running track, and distance). Through performing statistical analyses, including the distribution of HRV during the flat, straight segments and P-P plots of modified HRV, a driving workload calculation model was proposed. Integrating driving workload scales with values, the threshold of each scale of driving workload was determined by classification and regression tree (CART) algorithms.Results: The driving workload calculation model was suitable for driving speeds in the range of 40 to 120 km/h. The experimental data of 72 participants revealed that driving workload had a significant effect on modified HRV, revealing a change in driving speed. When the driving speed was between 100 and 120 km/h, drivers showed an apparent increase in the corresponding modified HRV. The threshold value of the normal driving workload K was between -0.0011 and 0.056 for a car driver and between -0.00086 and 0.067 for a truck driver.Conclusion: Heart rate variability was a direct and effective index for measuring driving workload despite being affected by multiple highway alignment elements. The driving workload model and the thresholds of driving workload classifications can be used to evaluate the quality of highway geometric design. A higher quality of highway geometric design could keep driving workload within a safer and more comfortable range. This study provided insight into reducing traffic injuries from the perspective of disciplinary integration of highway engineering and human factor engineering. [ABSTRACT FROM AUTHOR]

The operational land imager (OLI) is the latest instrument in the Landsat series of satellite imagery, which officially began normal operations on 30 May 2013. The OLI includes two bands that are not on the thematic mapper series of sensors aboard Landsat-5 and 7; a cirrus band and a coastal/aerosol band. This paper compares the classification and regression tree and the kernel-based extreme learning machine (KELM) for mapping crops in Hokkaido, Japan, using OLI data, except the cirrus band and the pan band. The OLI data acquired on 8 July 2013 was used for crop classification of beans, beets, grassland, maize, potatoes and winter wheat. The KELM algorithm performed better in this study and achieved overall accuracies of 90.1%. According to the Jeffries–Matusita (J–M) distances, the short wavelength infrared band provides the greater contribution (the highest value was observed for band 6 in OLI data). [ABSTRACT FROM PUBLISHER]

Abstract—This article presents a fault-classification method for transmission lines based on voltage phasors using classification and regression trees. The proposed method is intended to aid system operators in understanding the outputs of a phasor measurement unit only state estimator. Faults are classified into four categories when the estimator is positive sequence and into ten categories when the estimator is three phase. The fault data are generated in PowerWorld® (PowerWorld Corporation, Champaign, IL, USA) and DSA Tools® (Powertech Labs Inc., Surrey, British Columbia, Canada). The pre-fault state consists of a variety of operating conditions and loading angles of faulted lines. The fault condition comprises different fault types, fault locations, fault impedances, and fault incidence angles. Fault classification is done using MATLAB® (The MathWorks, Natick, Massachusetts, USA).The approach is successfully tested on the IEEE-118 bus system. The results demonstrate that the technique developed here is effective and robust, irrespective of the pre-fault and fault conditions. [ABSTRACT FROM PUBLISHER]

Yield improvement is an important issue in semiconductor-manufacturing supply chains, including color filter and micro-lens manufacturing. In the color filter and micro-lens processes, it is critical to quickly identify the defect pattern through the defect pictures and then take corrective actions to avoid greater yield loss. Until now, defect pattern recognition heavily relies on domain experts' knowledge, which easily causes inconsistent classification results from person to person and unsatisfactory performance. In this study, a framework is proposed to achieve automatic defect detection and classification in color filter and micro-lens manufacturing to enhance the decision quality of pattern recognition. In particular, the proposed framework integrates Canny edge detection and classification and regression tree methodology. To validate the viability of the proposed framework in real settings, an empirical study was conducted in collaboration with a leading complementary metal oxide semiconductor image sensor foundry in Taiwan. The results not only showed the effectiveness of the proposed framework but also demonstrated the practical values. [ABSTRACT FROM AUTHOR]

The aim of this study is to examine cruise passengers' characteristics, preferences and their overall experience in a port of call. Based on 1,361 survey data collected from passengers in the port of call of Cartagena, during the third quarter of 2009, a three-step multivariate market segment analysis is employed. First, a correspondence analysis is run to reveal the underlying factors in the data; second, based on the correspondence analysis, a hierarchical cluster investigation is performed to segment the sample into homogeneous groups; third, a decision tree is computed to characterise each group. The cluster analysis identifies six distinct market segments differentiated by nationality, satisfaction, safety perception and expenditure. The findings imply several policy directions. In particular, institutions should enhance the perception of safety in Cartagena to guarantee repeated visits, an ad hoc marketing policy may encourage revisit by young South Americans, and managers should extend the inland visiting time that is likely to produce local multiplier effects. [ABSTRACT FROM PUBLISHER]

This article discusses the successful implementation of Six Sigma methodology in a high precision and critical process in the manufacture of automotive products. The Six Sigma define–measure–analyse–improve–control approach resulted in a reduction of tolerance-related problems and improved the first pass yield from 85% to 99.4%. Data were collected on all possible causes and regression analysis, hypothesis testing, Taguchi methods, classification and regression tree, etc. were used to analyse the data and draw conclusions. Implementation of Six Sigma methodology had a significant financial impact on the profitability of the company. An approximate saving of US$70,000 per annum was reported, which is in addition to the customer-facing benefits of improved quality on returns and sales. The project also had the benefit of allowing the company to learn useful messages that will guide future Six Sigma activities. [ABSTRACT FROM AUTHOR]

A quantitative structure-retention relationship (QSRR) was developed to predict the retention of some pesticides in reversed-phase high-performance liquid chromatography with different mobile phase composition. A set of 1497 zero-to three-dimensional descriptors were used for each molecule in the data set. Classification and regression tree (CART) was successfully used as a descriptor selection method. In addition to mobile phase composition, two very simple descriptive descriptors (octanol-water partition coefficient and average molecular weight) were also selected and used as inputs for adaptive neuro-fuzzy inference system (ANFIS). The root mean square errors for the training, validation, and test sets were 0.027, 0.034, and 0.036, respectively. In comparison with artificial neural network method, the results showed CART-ANFIS is a powerful model for prediction of retention of pesticides. [ABSTRACT FROM PUBLISHER]

This article is concerned with the application of thresholding to the estimation of possibly sparse single sequence data observed subject to noise. In such problems, accuracy can be greatly improved by selecting a threshold that adapts to the unknown signal strength. We set out a classification and regression tree approach aimed at partitioning a sequence of inhomogeneous strength into component homogeneous regions where we can independently set a locally adaptive threshold and thus improve estimation. Our method places a mixture prior on each coefficient consisting of an atom of probability at zero and a symmetric probability density. The mixing weight is chosen via Empirical Bayes. The decision on whether a split should occur is based on a score test. Having selected the partitioning and obtained the local mixing weight for each region, estimation is carried out using the posterior median. We evaluate the performance of our method in the single sequence case and for wavelet denoising on both simulated and real data. In the wavelet context we consider two alternative implementations, splitting the coefficients levelwise and splitting the original domain. Our method is cheap to compute and in numerical comparisons our method shows excellent performance when compared with current thresholding techniques. This article has supplementary material online. [ABSTRACT FROM AUTHOR]

Background and purpose: Residual brain function has been documented in vegetative state patients, yet early prognosis remains difficult. The purpose of this study was to identify by artificial intelligence procedures (classification and regression trees, data-mining) the significant neurological signs correlated to and predictive of outcome. Methods: Three hundred and thirty-three patients in vegetative state of traumatic or non-traumatic aetiology referred to the S.Anna Institute were retrospectively studied. Twenty-two neurological signs were assessed according to criteria included in the UNI ENI ISO 9001 : 2000 quality standards at admission (Time0) and after 50, 100 or 180 days and entered into a CART (classification and regression tree) data-mining procedure with a decisional tree j48 (Weka software and 10-fold cross-validation). Outcome was conventionally rated by the Glasgow outcome scale. Results and conclusions: Re-appearance with proper timing of spontaneous motility, eye tracking and oculo-cephalic reflex and disappearance of oral automatisms proved highly correlated to outcome and allowed early and reliable prognosis. These findings are consistent with the brain functional organization thought to sustain consciousness and warrant systematic investigation. Classification and regression trees and data-mining procedures proved applicable in neurology to sort out significant clinical signs also in clinical conditions characterized by paucity of signs such as the vegetative state. Extended application in clinical medicine is conceivable based on the approach peculiarities. [ABSTRACT FROM AUTHOR]

An optimal engineering design problem is challenging because nonlinear objective functions usually need to be evaluated in a high-dimensional design space. This article presents a data-mining-aided optimal design method, that is able to find a competitive design solution with a relatively low computational cost. The method consists of four components: (1) a uniform-coverage selection method, that chooses design representatives from among a large number of original design alternatives for a nonrectangular design space; (2) feature functions, of which evaluation is computationally economical as the surrogate for the design objective function; (3) a clustering method, that generates a design library based on the evaluation of feature functions instead of an objective function; and (4) a classification method to create the design selection rules, eventually leading us to a competitive design. Those components are implemented to facilitate the optimal fixture layout design in a multistation panel assembly process. The benefit of the data-mining-aided optimal design is clearly demonstrated by comparison with both local optimization methods (e.g., simplex search) and random search-based optimizations (e.g., simulated annealing). [ABSTRACT FROM AUTHOR]