Classification And Regression Tree Approach Research Articles

Identifying the sociodemographic and work-related factors related to workers’ daily physical activity using a decision tree approach

BackgroundThe social and behavioural factors related to physical activity among adults are well known. Despite the overlapping nature of these factors, few studies have examined how multiple predictors of physical activity interact. This study aimed to identify the relative importance of multiple interacting sociodemographic and work-related factors associated with the daily physical activity patterns of a population-based sample of workers.MethodsSociodemographic, work, screen time, and health variables were obtained from five, repeated cross-sectional cohorts of workers from the Canadian Health Measures Survey (2007 to 2017). Classification and Regression Tree (CART) modelling was used to identify the discriminators associated with six daily physical activity patterns. The performance of the CART approach was compared to a stepwise multinomial logistic regression model.ResultsAmong the 8,909 workers analysed, the most important CART discriminators of daily physical activity patterns were age, job skill, and physical strength requirements of the job. Other important factors included participants’ sex, educational attainment, fruit/vegetable intake, industry, work hours, marital status, having a child living at home, computer time, and household income. The CART tree had moderate classification accuracy and performed marginally better than the stepwise multinomial logistic regression model.ConclusionAge and work-related factors–particularly job skill, and physical strength requirements at work–appeared as the most important factors related to physical activity attainment, and differed based on sex, work hours, and industry. Delineating the hierarchy of factors associated with daily physical activity may assist in targeting preventive strategies aimed at promoting physical activity in workers.

Classification trees for analysing highly educated people satisfaction with airlines’ services

Passengers' satisfaction has relevant implications in air transport system. This study wants to provide further insights in this field in order to give managers and practitioners simple and explanatory results that allow them to increase passengers' satisfaction. The paper explores passengers' satisfaction with airline services through the Classification and Regression Tree (CART) approach together with the Importance Performance Analysis (IPA). CART approach allows identifying the relevance of each service factor affecting user's satisfaction, while IPA merges the CART results with the satisfaction directly expressed by the passengers in order to identify the most critical factors. A sample of passengers travelling by direct flights has been used. The results show that passengers have great expectations in terms of staff competence and courtesy, while seem less concerned about comfort on board. In addition, as it emerges from the IPA, some services, such as boarding and landing operations, should be improved.

An Interpretable Machine Learning Model to Classify Coronary Bifurcation Lesions.

Coronary bifurcation lesions are a leading cause of Coronary Artery Disease (CAD). Despite its prevalence, coronary bifurcation lesions remain difficult to treat due to our incomplete understanding of how various features of lesion anatomy synergistically disrupt normal hemodynamic flow. In this work, we employ an interpretable machine learning algorithm, the Classification and Regression Tree (CART), to model the impact of these geometric features on local hemodynamic quantities. We generate a synthetic arterial database via computational fluid dynamic simulations and apply the CART approach to predict the time averaged wall shear stress (TAWSS) at two different locations within the cardiac vasculature. Our experimental results show that CART can estimate a simple, interpretable, yet accurately predictive nonlinear model of TAWSS as a function of such features.Clinical relevance- The fitted tree models have the potential to refine predictions of disturbed hemodynamic flow based on an individual's cardiac and lesion anatomy and consequently makes progress towards personalized treatment planning for CAD patients.

Multimodal Approach to Predict Neurological Outcome after Cardiac Arrest: A Single-Center Experience.

Introduction: The aims of this study were to assess the concordance of different tools and to describe the accuracy of a multimodal approach to predict unfavorable neurological outcome (UO) in cardiac arrest patients. Methods: Retrospective study of adult (>18 years) cardiac arrest patients who underwent multimodal monitoring; UO was defined as cerebral performance category 3–5 at 3 months. Predictors of UO were neurological pupillary index (NPi) ≤ 2 at 24 h; highly malignant patterns on EEG (HMp) within 48 h; bilateral absence of N20 waves on somato-sensory evoked potentials; and neuron-specific enolase (NSE) > 75 μg/L. Time-dependent decisional tree (i.e., NPi on day 1; HMp on day 1–2; absent N20 on day 2–3; highest NSE) and classification and regression tree (CART) analysis were used to assess the prediction of UO. Results: Of 137 patients, 104 (73%) had UO. Abnormal NPi, HMp on day 1 or 2, the bilateral absence of N20 or NSE >75 mcg/L had a specificity of 100% to predict UO. The presence of abnormal NPi was highly concordant with HMp and high NSE, and absence of N20 or high NSE with HMp. However, HMp had weak to moderate concordance with other predictors. The time-dependent decisional tree approach identified 73/103 patients (70%) with UO, showing a sensitivity of 71% and a specificity of 100%. Using the CART approach, HMp on EEG was the only variable significantly associated with UO. Conclusions: This study suggests that patients with UO had often at least two predictors of UO, except for HMp. A multimodal time-dependent approach may be helpful in the prediction of UO after CA. EEG should be included in all multimodal prognostic models.

Measuring the Response Performance of U.S. States against COVID-19 Using an Integrated DEA, CART, and Logistic Regression Approach.

Measuring the U.S.’s COVID-19 response performance is an extremely important challenge for health care policymakers. This study integrates Data Envelopment Analysis (DEA) with four different machine learning (ML) techniques to assess the efficiency and evaluate the U.S.’s COVID-19 response performance. First, DEA is applied to measure the efficiency of fifty U.S. states considering four inputs: number of tested, public funding, number of health care employees, number of hospital beds. Then, number of recovered from COVID-19 as a desirable output and number of confirmed COVID-19 cases as a undesirable output are considered. In the second stage, Classification and Regression Tree (CART), Boosted Tree (BT), Random Forest (RF), and Logistic Regression (LR) were applied to predict the COVID-19 response performance based on fifteen environmental factors, which were classified into social distancing, health policy, and socioeconomic measures. The results showed that 23 states were efficient with an average efficiency score of 0.97. Furthermore, BT and RF models produced the best prediction results and CART performed better than LR. Lastly, urban, physical inactivity, number of tested per population, population density, and total hospital beds per population were the most influential factors on efficiency.

Underweight in Nursing Homes: Differences between Men and Women

Objective: In Germany, there is an ongoing concern about the high prevalence of underweight on admission to health-care institutions. In order to assess possible sex-specific differences, the aim of this study is to provide valid figures about the prevalence and risk factors of underweight of men and women in German nursing homes. Material and Methods: A secondary data analysis of 8 annual consecutive cross-sectional studies of 19,686 residents from 280 nursing homes was conducted from 2009 to 2016. Underweight was defined as BMI < 18.5 (<20) for individuals <65 years (≥65 years). For statistical modeling, we used classification and regression trees (CRTs) and random forest in “R.” Results: Average prevalence of underweight in nursing home residents was 13.7% (13.2–14.2). Initial descriptive results showed that the prevalence of underweight among women was 15.6% (15.0–16.2) and the prevalence of underweight among men was 7.5% (6.7–8.2). The CRT-based modeling indicated that “loss of appetite” as the most important indicator for low BMI. If “loss of appetite” was present, prevalence of underweight increased from 13.5 to 39.1%. Other important indicators were “very large institutions” and the “resident/nurse ratio.” The random forest analysis confirmed the importance of the CRT approach. Discussion/Conclusion: The multivariate approach revealed that the role of sex for being underweight in nursing homes is marginal. To avoid higher morbidity and mortality in this group, nutritional intervention by clinical practitioners to increase appetite should be given high priority, especially in large long-term care institutions.

Development and validation of a Brief Diet Quality Assessment Tool in the French-speaking adults from Quebec

BackgroundThe objective of this study was to develop and validate a short, self-administered questionnaire to assess diet quality in clinical settings, using the Alternative Healthy Eating Index (AHEI) as reference.MethodsA total of 1040 men and women (aged 44.6 ± 14.4 y) completed a validated web-based food frequency questionnaire (webFFQ) and had their height and weight measured (development sample). Participants were categorized arbitrarily according to diet quality (high: AHEI score ≥ 65/110, low: AHEI score < 65/110) based on dietary intake data from the webFFQ. The Brief Diet Quality Assessment Tool was developed using a classification and regression tree (CART) approach and individual answers to the webFFQ among participants considered to have a plausible energy intake (ratio of reported energy intake to basal metabolic rate ≥ 1.2 and < 2.4; n = 1040). A second sample of 3344 older adults (aged 66.5 ± 6.4 y) was used to test the external validity of the Brief Diet Quality Assessment Tool (external validation sample).ResultsThe decision tree included sequences of 3 to 6 binary questions, yielding 21 different pathways classifying diet quality as being high or low. In the development sample, the area under the receiver operating characteristic (ROC) curve of the predictive model was 0.92, with sensitivity, specificity and agreement values of 89.5, 83.9 and 87.2%. Compared with individuals having a low-quality diet according to the Brief Diet Quality Assessment Tool (mean AHEI 56.7 ± 11.4), individuals classified as having a high-quality diet (mean AHEI 71.3 ± 11.0) were significantly older, and had lower BMI, percent body fat and waist circumference, and had lower blood pressure, triglycerides, cholesterol/HDL ratio and fasting insulin as well as higher HDL-cholesterol concentrations (all P < 0.05). Similar results were observed in the external validation sample, although overall performance of the Brief Diet Quality Assessment Tool was slightly lower than in the development sample, with an area under the ROC curve of 0.79 and sensitivity, specificity and agreement values of 73.0, 69.0 and 71.3%, respectively.ConclusionThe CART approach yielded a simple and rapid Brief Diet Quality Assessment Tool that identifies individuals at risk of having a low-quality diet. Further studies are needed to test the performance of this tool in primary care settings.

Investigation of helmet use behavior of motorcyclists and effectiveness of enforcement campaign using CART approach

Use of the helmet is beneficial in reducing the severity of injuries and avoiding fatalities for motorcyclists, therefore, legislation exists in almost all countries. In practice, the situation is different regarding the helmet use, especially in the developing countries. This paper investigates the helmet use behavior of motorcyclists in Karachi, Pakistan. It determines the significant factors affecting the helmet use in Karachi (Pakistan) and recommending effective campaign measures to promote helmet use. It is vital in relation to the exponential growth of motorcycle users and poor enforcement of traffic rules. Repeated cross-sectional data collected before (n = 226) and after (n = 277) the helmet use enforcement campaign is analyzed using univariate and non-parametric classification and regression tree (CART) techniques. A significant number of motorcyclists do not hold a driving license and CART analysis highlighted the significance of this variable along with exposure to the road environment (measured as daily usage of motorcycle) to explain helmet use. Campaign effects are found temporary, therefore, serious efforts are required to design coherent and structured awareness and enforcement programs.

Predicting Renal Recovery After Dialysis-Requiring Acute Kidney Injury

IntroductionAfter dialysis-requiring acute kidney injury (AKI-D), recovery of sufficient kidney function to discontinue dialysis is an important clinical and patient-oriented outcome. Predicting the probability of recovery in individual patients is a common dilemma.MethodsThis cohort study examined all adult members of Kaiser Permanente Northern California who experienced AKI-D between January 2009 and September 2015 and had predicted inpatient mortality of <20%. Candidate predictors included demographic characteristics, comorbidities, laboratory values, and medication use. We used logistic regression and classification and regression tree (CART) approaches to develop and cross-validate prediction models for recovery.ResultsAmong 2214 patients with AKI-D, mean age was 67.1 years, 40.8% were women, and 54.0% were white; 40.9% of patients recovered. Patients who recovered were younger, had higher baseline estimated glomerular filtration rates (eGFR) and preadmission hemoglobin levels, and were less likely to have prior heart failure or chronic liver disease. Stepwise logistic regression applied to bootstrapped samples identified baseline eGFR, preadmission hemoglobin level, chronic liver disease, and age as the predictors most commonly associated with coming off dialysis within 90 days. Our final logistic regression model including these predictors had a correlation coefficient between observed and predicted probabilities of 0.97, with a c-index of 0.64. An alternate CART approach did not outperform the logistic regression model (c-index 0.61).ConclusionWe developed and cross-validated a parsimonious prediction model for recovery after AKI-D with excellent calibration using routinely available clinical data. However, the model’s modest discrimination limits its clinical utility. Further research is needed to develop better prediction tools.

Classification and Regression Tree Approach for Prediction of Potential Hazards of Urban Airborne Bacteria during Asian Dust Events

Despite progress in monitoring and modeling Asian dust (AD) events, real-time public hazard prediction based on biological evidence during AD events remains a challenge. Herein, both a classification and regression tree (CART) and multiple linear regression (MLR) were applied to assess the applicability of prediction for potential urban airborne bacterial hazards during AD events using metagenomic analysis and real-time qPCR. In the present work, Bacillus cereus was screened as a potential pathogenic candidate and positively correlated with PM10 concentration (p < 0.05). Additionally, detection of the bceT gene with qPCR, which codes for an enterotoxin in B. cereus, was significantly increased during AD events (p < 0.05). The CART approach more successfully predicted potential airborne bacterial hazards with a relatively high coefficient of determination (R2) and small bias, with the smallest root mean square error (RMSE) and mean absolute error (MAE) compared to the MLR approach. Regression tree analyses from the CART model showed that the PM10 concentration, from 78.4 µg/m3 to 92.2 µg/m3, is an important atmospheric parameter that significantly affects the potential airborne bacterial hazard during AD events. The results show that the CART approach may be useful to effectively derive a predictive understanding of potential airborne bacterial hazards during AD events and thus has a possible for improving decision-making tools for environmental policies associated with air pollution and public health.

Precipitation forecasting using classification and regression trees (CART) model: a comparative study of different approaches

Interest in semiarid climate forecasting has prominently grown due to risks associated with above average levels of precipitation amount. Longer-lead forecasts in semiarid watersheds are difficult to make due to short-term extremes and data scarcity. The current research is a new application of classification and regression trees (CART) model, which is rule-based algorithm, for prediction of the precipitation over a highly complex semiarid climate system using climate signals. We also aimed to compare the accuracy of the CART model with two most commonly applied models including time series modeling (ARIMA), and adaptive neuro-fuzzy inference system (ANFIS) for prediction of the precipitation. Various combinations of large-scale climate signals were considered as inputs. The results indicated that the CART model had a better results (with Nash–Sutcliffe efficiency, NSE > 0.75) compared to the ANFIS and ARIMA in forecasting precipitation. Also, the results demonstrated that the ANFIS method can predict the precipitation values more accurately than the time series model based on various performance criteria. Further, fall forecasts ranked “very good” for the CART method, while the ANFIS and the time series model approximately indicated “satisfactory” and “unsatisfactory” performances for all stations, respectively. The forecasts from the CART approach can be helpful and critical for decision makers when precipitation forecast heralds a prolonged drought or flash flood.

Analyzing Monte Carlo Simulation Studies With Classification and Regression Trees

Monte Carlo simulations are an important tool for researchers to study statistical properties of estimators, such as parameter bias, or the limits of various modeling approaches. Typically, the immense amount of data produced by Monte Carlo studies is analyzed with regression or analysis of variance, and researchers are faced with making arbitrary decisions regarding what effects to report and what interactions to test. Understanding current limitations, we propose a classification and regression trees (CART) approach from the statistical learning and data mining field to analyze Monte Carlo simulation data. We demonstrate the advantages of the CART approach and several extensions by reanalyzing and interpreting results from one published Monte Carlo study and one fully reproducible simulation example. Results suggest that CART is able to arrive at the same conclusions as current descriptive and inferential approaches and, at the same time, provide additional insight on the complex interactions among simulation factors.

Is atmospheric pollution exposure during pregnancy associated with individual and contextual characteristics? A nationwide study in France

BackgroundExposure to atmospheric pollutants is a danger for the health of pregnant mother and children. Our objective was to identify individual (socioeconomic and behavioural) and contextual factors associated with atmospheric...

Information Mining from Heterogeneous Data Sources: A Case Study on Drought Predictions

The objective of this study was to develop information mining methodology for drought modeling and predictions using historical records of climate, satellite, environmental, and oceanic data. The classification and regression tree (CART) approach was used for extracting drought episodes at different time-lag prediction intervals. Using the CART approach, a number of successful model trees were constructed, which can easily be interpreted and used by decision makers in their drought management decisions. The regression rules produced by CART were found to have correlation coefficients from 0.71–0.95 in rules-alone modeling. The accuracies of the models were found to be higher in the instance and rules model (0.77–0.96) compared to the rules-alone model. From the experimental analysis, it was concluded that different combinations of the nearest neighbor and committee models significantly increase the performances of CART drought models. For more robust results from the developed methodology, it is recommended that future research focus on selecting relevant attributes for slow-onset drought episode identification and prediction.

Developing an Adaptive Treatment Strategy for Peer-Related Social Skills for Children With Autism Spectrum Disorders

The purpose of this study was to understand the trajectories of children's response to an intervention prior to the end of the treatment in order to inform adaptive treatment models for future studies. Participants with autism spectrum disorder (ASD) were drawn from a randomized controlled trial comparing 2 different social skills interventions at children's schools. We excluded children with ASD who entered the study with at least 80% time engaged (the average time of neurotypical children in the same classes) in order to examine only those who were engaged below the typical developing peers' average percentage of time engaged. The final sample included 92 children with ASD (82% male, average age = 8.14 years, average IQ = 89.6). We explored whether playground engagement scores measured at entry and midpoint of treatment predicted their engagement scores at end of treatment using the Classification and Regression Tree (CART) method. Using the CART approach, 4 meaningful subgroups based on children's playground engagement scores measured at entry and changes from entry to midpoint were identified. These data suggest that measurements of children's behavior midstudy can be used to predict children's treatment outcomes. Such data may be used to inform decisions to augment or alter programming prior to treatment end in order to tailor intervention to best meet the needs of individual children.

Assessing the Predictive Utility of Logistic Regression, Classification and Regression Tree, Chi-Squared Automatic Interaction Detection, and Neural Network Models in Predicting Inmate Misconduct

This study assesses the relative utility of a traditional regression approach - logistic regression (LR) - and three classification techniques - classification and regression tree (CART), chi-squared automatic interaction detection (CHAID), and multi-layer perceptron neural network (MLPNN)—in predicting inmate misconduct. The four models were tested using a sample of inmates held in state and federal prisons and predictors derived from the importation model on inmate adaptation. Multi-validation procedure and multiple evaluation indicators were used to evaluate and report the predictive accuracy. The overall accuracy of the four models varied between 0.60 and 0.66 with an overall AUC range of 0.60–0.70. The LR and MLPNN methods performed significantly better than the CART and CHAID techniques at identifying misbehaving inmates and the CHAID method outperformed the CART approach in classifying defied inmates. The MLPNN method performed significantly better than the LR technique in predicting inmate misconduct among the training samples.

The benefits of tree-based models for stock selection

The performance of stock relative to its peer group is influenced by a multitude of factors and their interactions, which are typically modelled by investment practitioners in a classical parametric framework. Although such models are in many cases useful for identifying linear interactions, they are less well suited to capturing the higher-order relationships between a company's fundamental characteristics and its subsequent relative return. Despite this, non-parametric and non-linear approaches such as classification and regression trees (CART) have been largely overlooked by the finance industry, which still relies heavily upon linear factor models. This article investigates the use of CART for stock selection within North America in order to highlight some of the advantages of adopting a broader suite of modelling tools. Its focus is on the period since the onset of the Global Financial Crisis in 2007 to late 2010 – a period associated with elevated volatility and sharp swings in investor sentiment. More specifically, we directly compare a CART model against a more traditional linear framework. We observe that the performance of portfolios formed from a tree-based model was quite robust during both the 2007/2008 downturn in equities and the subsequent market recovery. As such, we believe that stock selection models based on the CART approach offer an attractive opportunity to diversify model risk.

Classification and Regression Tree Approach for Predicting Drivers’ Merging Behavior in Short-Term Work Zone Merging Areas

This study aims to use the classification and regression tree (CART) approach, one of the most powerful data mining techniques, to predict drivers’ merging behavior in a work zone merging area. On the basis of the eight factors affecting drivers’ merging behavior, a binary CART is built using the merging traffic data collected from a short-term work zone site in Singapore. The CART comprises 7 levels and 15 leaf nodes to predict drivers’ merging behavior in the work zone merging area. The results show that the CART provides much higher prediction accuracy than the conventional binary logit model. Traffic engineers can easily understand how drivers make merging/nonmerging decisions. This demonstrates that the CART approach is a good alternative for investigating drivers’ merging behavior in work zone merging areas.

Aligned Collagen Is a Prognostic Signature for Survival in Human Breast Carcinoma

Evidence for the potent influence of stromal organization and function on invasion and metastasis of breast tumors is ever growing. We have performed a rigorous examination of the relationship of a tumor-associated collagen signature-3 (TACS-3) to the long-term survival rate of human patients. TACS-3 is characterized by bundles of straightened and aligned collagen fibers that are oriented perpendicular to the tumor boundary. An evaluation of TACS-3 was performed in biopsied tissue sections from 196 patients by second harmonic generation imaging of the backscattered signal generated by collagen. Univariate analysis of a Cox proportional hazard model demonstrated that the presence of TACS-3 was associated with poor disease-specific and disease-free survival, resulting in hazard ratios between 3.0 and 3.9. Furthermore, TACS-3 was confirmed to be an independent prognostic indicator regardless of tumor grade and size, estrogen or progesterone receptor status, human epidermal growth factor receptor-2 status, node status, and tumor subtype. Interestingly, TACS-3 was positively correlated to expression of stromal syndecan-1, a receptor for several extracellular matrix proteins including collagens. Because of the strong statistical evidence for poor survival in patients with TACS, and because the assessment can be performed in routine histopathological samples imaged via second harmonic generation or using picrosirius, we propose that quantifying collagen alignment is a viable, novel paradigm for the prediction of human breast cancer survival.

Improved Land Cover Mapping using Random Forests Combined with Landsat Thematic Mapper Imagery and Ancillary Geographic Data

Large area land-cover mapping involving large volumes of data is becoming more common in remote sensing applications. Thus, there is a pressing need for increased automation in the land-cover mapping process. The main objective of this research was to compare the performance of three machine learning algorithms (MLAS) for mapping wetlands in the Sanjiang Plain combined Landsat TM imagery with ancillary geographical data. Three MLAS included random forest (RF), classification and regression tree (CART), and maximum likelihood classification (MLC). Comparisons were based on several criteria: overall accuracy, sensitivity to data set size, and noise. Our results indicated that first, the random forest and CART approach can achieve substantial improvements in accuracy over the traditional MLC method. Random forest produced the highest overall accuracy (91.3 percent) the kappa coefficient 0.8943, with marsh class accuracies ranging from 77.4 percent to 90.0 percent. Secondly, the random forest method was least sensitive to reduction in training sample size, and it was most resistant to the presence of noise compared to CART and MLC. The comparison between three MLAS revealed that the random forest approach was most resistant to training data deficiencies while improved land-cover map accuracy in marsh area.