Using machine learning to select variables in data envelopment analysis: Simulations and application using electricity distribution data

The present study aimed to rank banks in terms of board of directors report and notifying the users of reports. In addition, we evaluated factors affecting risk disclosure from the perspective of corporate governance. Moreover, we assessed risk disclosure based on linguistic analysis of board report text and capital adequacy ratio. Words were applied as measurement units to measure risk disclosure. The advantage of this technique is the unique analysis of words. According to the theoretical foundations presented in the present study, we first identified risk disclosure words in reports provided to financial information users and divided them into two categories of positive and negative risk disclosure words. Another variable selected for risk disclosure was capital adequacy ratio. Effective variables in corporate governance system in banks included the board independence, duality of CEO duties, and major shareholders as input variables in data envelopment analysis (DEA) model. On the other hand, the BCC model of DEA was selected as output-based nature. The statistical population included all banks listed in Tehran Stock Exchange. In total, 20 year-bank units listed during 2016-2017 were assessed. In the end, seven year-bank units were considered efficient while the rest were inefficient. Moreover, we estimated the amount of shortage in outputs to reach inefficient banks to the desired level of efficiency.

Stepwise selection of variables in data envelopment analysis: Procedures and managerial perspectives

We show that the adaptive Lasso (aLasso) and the adaptive group Lasso (agLasso) are oracle efficient in stationary vector autoregressions where the number of parameters per equation is smaller than the number of observations. In particular, this means that the parameters are estimated consistently at a √T-rate, that the truly zero parameters are classified as such asymptotically, and that the non-zero parameters are estimated as efficiently as if only the relevant variables had been included in the model from the outset. The group adaptive Lasso differs from the adaptive Lasso by dividing the covariates into groups whose members are all relevant or all irrelevant. Both estimators have the property that they perform variable selection and estimation in one step. We evaluate the forecasting accuracy of these estimators for a large set of macroeconomic variables. The plain Lasso is found to be the most precise procedure overall. The adaptive and the adaptive group Lasso are less stable but mostly perform at par with common factor models.

Immune checkpoint inhibitor-related pneumonitis (ICIP) is a common and potentially life-threatening adverse event with non-specific symptoms. It is of significance to identify high-risk population of ICIP. However, existing prediction models for ICIP are often limited by their reliance on clinically inaccessible variables and homogeneous methodologies, hindering their clinical utility. This study aimed to develop a clinical risk-prediction model for ICIP in patients with gastrointestinal (GI) cancer based on four machine learning (ML) methods. We conducted a retrospective analysis of data from GI cancer patients who received immune checkpoint inhibitors (ICIs) between 2018 and 2022 in Beijing Cancer Hospital. For each patient, 36 clinical indicators associated with pneumonia risk were gathered. The dataset was split into training and testing sets in a ratio of 7:3. Variable selection was first performed using Least Absolute Shrinkage and Selection Operator (LASSO) regression. Subsequently, four ML algorithms: logistic regression (LR), random forest (RF), Support vector machine (SVM), and Adaptive Boosting (AdaBoost), were employed to develop and validate ICIP prediction models. The models' performance was assessed using sensitivity, specificity, precision, F1-score, and the area under the receiver operating characteristic curve (AUC) value. The optimal cutoff point for the best model was determined and a web-based tool was developed based on it. We collected medical data from 1,101 GI cancer patients. Ten predictive variables were identified as significant: gender, age, treatment line, smoking index, drinking history, lung metastasis, neutrophil-to-lymphocyte ratio, platelet-to-lymphocyte ratio, hemoglobin, and albumin. After constructing and comparing four ML models, the RF model demonstrated best performance with an AUC of 0.899. The web-based tool for ICIP risk prediction is available at https://healthy.aistarfish.com/business/pneumonia-prediction/#/home. We analyzed 36 clinical predictors of ICIP in 1,101 patients treated with ICIs, and 10 variables were included. The smoking index, albumin and hemoglobin emerged as novel predictors specific to GI cancers. Among the models constructed using four ML methods, the RF model showed the best performance. Additionally, a web-based tool was developed to facilitate the early clinical identification of populations at high risk of ICIP. Future directions include external validation of the model to enhance clinical usability.

The appreciative democratic voice of DEA: A case of faculty academic performance evaluation

In the field of quantitative remote sensing of forest biomass, a prominent phenomenon is the increasing number of explanatory variables. Then how to effectively select explanatory variables has become an important issue. Linear regression model is one of the commonly used remote sensing models. In the process of establishing the linear regression model, a vital step is to select explanatory variables. Focusing on variable selection and model stability, this paper conducts a comparative study on the performance of eight linear regression parameter estimation methods (Stepwise Regression Method (SR), Criterions Based on The Bayes Method (BIC), Criterions Based on The Bayes Method (AIC), Criterions Based on Prediction Error (Cp), Least Absolute Shrinkage and Selection Operator (Lasso), Adaptive Lasso, Smoothly Clipped Absolute Deviation (SCAD), Non-negative garrote (NNG)) in the subtropical forest biomass remote sensing model development. For the purpose of comparison, OLS and RR, are commonly used as methods with no variable selection ability, and are also compared and discussed. The performance of five aspects are evaluated in this paper: (i) Determination coefficient, prediction error, model error, etc., (ii) significance test about the difference between determination coefficients, (iii) parameter stability, (iv) variable selection stability and (v) variable selection ability of the methods. All the results are obtained through a five ten-fold CV. Some evaluation indexes are calculated with or without degrees of freedom. The results show that BIC performs best in comprehensive evaluation, while NNG, Cp and AIC perform poorly as a whole. Other methods show a great difference in the performance on each index. SR has a strong capability in variable selection, although it is poor in commonly used indexes. The short-wave infrared band and the texture features derived from it are selected most frequently by various methods, indicating that these variables play an important role in forest biomass estimation. Some of the conclusions in this paper are likely to change as the study object changes. The ultimate goal of this paper is to introduce various model establishment methods with variable selection capability, so that we can have more choices when establishing similar models, and we can know how to select the most appropriate and effective method for specific problems.

Meso-level Comparison of Mental Health Service Availability and Use in Chile and Spain

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.

Relevance. Public health effectiveness is crucial for population health, especially in the face of global challenges like the COVID-19 pandemic. The study applies the Data Envelopment Analysis (DEA) model to measure the efficiency of operating costs in the health care system across various regions of Kazakhstan from 2017 to 2021. Existing methods for assessing healthcare effectiveness often overlook the system's complexity, which turns DEA into a valuable tool to identify inequalities in healthcare availability and quality. Research objective. This study aims to employ the DEA model to measure the efficiency of operating costs in the health care system across regions of Kazakhstan in 2017-2021. Data and methods. The DEA model was chosen for its ability to analyze the efficiency of operating costs. Data were collected from the Agency for Strategic Planning and Reforms of the Republic of Kazakhstan. Results. Our findings indicate the need for increased healthcare financing in specific regions, emphasizing the importance of transparent spending. The study concludes that the DEA model can regularly assess health financing, ensuring resources are directed where most needed. The novelty lies in establishing a link between financing and health outcomes. Conclusions. The study's results and methodology can be used by public health authorities in assessing operating costs' effectiveness, allocating resources judiciously, and making informed decisions to enhance the healthcare system.

Data Envelopment Analysis (DEA) models and Free Disposal Hull (FDH) models were proposed almost simultaneously in the scientific literature. In DEA models, the constraints generate a convex set. For this reason, optimization methods and software are widely used for modeling and computations for these models. Non-convexity of production possibility set of FDH models refrained significantly the development of these models. In this paper, two methods are proposed for two-and three-dimensional frontier visualization for DEA and FDH models. Computational experiments using real-life datasets documented reliability and effectiveness of proposed methods.

Many tree algorithms have been developed for regression problems. Although they are regarded as good algorithms, most of them suffer from loss of prediction accuracy when there are many irrelevant variables and the number of predictors exceeds the number of observations. We propose the multistep regression tree with adaptive variable selection to handle this problem. The variable selection step and the fitting step comprise the multistep method. The multistep generalized unbiased interaction detection and estimation (GUIDE) with adaptive forward selection (fg) algorithm, as a variable selection tool, performs better than some of the well-known variable selection algorithms such as efficacy adaptive regression tube hunting (EARTH), FSR (false selection rate), LSCV (least squares cross-validation), and LASSO (least absolute shrinkage and selection operator) for the regression problem. The results based on simulation study show that fg outperforms other algorithms in terms of selection result and computation time. It generally selects the important variables correctly with relatively few irrelevant variables, which gives good prediction accuracy with less computation time.

Classifying 2-year recurrence in patients with dlbcl using clinical variables with imbalanced data and machine learning methods

Data envelopment analysis (DEA) is a non-parametric technique for measuring the efficiency of decision making units (DMU) with exact input and output terms. So far, a number of DEA models with interval data have been developed. The radial DEA model such as CCR or BCC model with interval data are well known as basic DEA models with interval data. However, these radial DEA models have some defects. In this study, we develop a new approach based upon the slacks-based measurement (SBM) of super-efficiency model for dealing with interval data in DEA. The SBM super-efficiency DEA model can not only discriminate the performance of efficient DMUs from inefficient ones, but can distinguish between the efficient DMUs. In addition, this model can overcome the drawback that the super-efficiency model can be infeasible under variable returns to scale (VRS) assumption. An illustrative example is provided to verify the idea of this paper and show its potential application and validity.

Purpose The purpose of this paper is to provide an outline of the major contributions in the literature on the determination of the least distance in data envelopment analysis (DEA). The focus herein is primarily on methodological developments. Specifically, attention is mainly paid to modeling aspects, computational features, the satisfaction of properties and duality. Finally, some promising avenues of future research on this topic are stated. Design/methodology/approach DEA is a methodology based on mathematical programming for the assessment of relative efficiency of a set of decision-making units (DMUs) that use several inputs to produce several outputs. DEA is classified in the literature as a non-parametric method because it does not assume a particular functional form for the underlying production function and presents, in this sense, some outstanding properties: the efficiency of firms may be evaluated independently on the market prices of the inputs used and outputs produced; it may be easily used with multiple inputs and outputs; a single score of efficiency for each assessed organization is obtained; this technique ranks organizations based on relative efficiency; and finally, it yields benchmarking information. DEA models provide both benchmarking information and efficiency scores for each of the evaluated units when it is applied to a dataset of observations and variables (inputs and outputs). Without a doubt, this benchmarking information gives DEA a distinct advantage over other efficiency methodologies, such as stochastic frontier analysis (SFA). Technical inefficiency is typically measured in DEA as the distance between the observed unit and a “benchmarking” target on the estimated piece-wise linear efficient frontier. The choice of this target is critical for assessing the potential performance of each DMU in the sample, as well as for providing information on how to increase its performance. However, traditional DEA models yield targets that are determined by the “furthest” efficient projection to the evaluated DMU. The projected point on the efficient frontier obtained as such may not be a representative projection for the judged unit, and consequently, some authors in the literature have suggested determining closest targets instead. The general argument behind this idea is that closer targets suggest directions of enhancement for the inputs and outputs of the inefficient units that may lead them to the efficiency with less effort. Indeed, authors like Aparicio et al. (2007) have shown, in an application on airlines, that it is possible to find substantial differences between the targets provided by applying the criterion used by the traditional DEA models, and those obtained when the criterion of closeness is utilized for determining projection points on the efficient frontier. The determination of closest targets is connected to the calculation of the least distance from the evaluated unit to the efficient frontier of the reference technology. In fact, the former is usually computed through solving mathematical programming models associated with minimizing some type of distance (e.g. Euclidean). In this particular respect, the main contribution in the literature is the paper by Briec (1998) on Hölder distance functions, where formally technical inefficiency to the “weakly” efficient frontier is defined through mathematical distances. Findings All the interesting features of the determination of closest targets from a benchmarking point of view have generated, in recent times, the increasing interest of researchers in the calculation of the least distance to evaluate technical inefficiency (Aparicio et al., 2014a). So, in this paper, we present a general classification of published contributions, mainly from a methodological perspective, and additionally, we indicate avenues for further research on this topic. The approaches that we cite in this paper differ in the way that the idea of similarity is made operative. Similarity is, in this sense, implemented as the closeness between the values of the inputs and/or outputs of the assessed units and those of the obtained projections on the frontier of the reference production possibility set. Similarity may be measured through multiple distances and efficiency measures. In turn, the aim is to globally minimize DEA model slacks to determine the closest efficient targets. However, as we will show later in the text, minimizing a mathematical distance in DEA is not an easy task, as it is equivalent to minimizing the distance to the complement of a polyhedral set, which is not a convex set. This complexity will justify the existence of different alternatives for solving these types of models. Originality/value As we are aware, this is the first survey in this topic.

Multiple imputation (MI) is a commonly used technique for handling missing data in large-scale medical and public health studies. However, variable selection on multiply-imputed data remains an important and longstanding statistical problem. If a variable selection method is applied to each imputed dataset separately, it may select different variables for different imputed datasets, which makes it difficult to interpret the final model or draw scientific conclusions. In this paper, we propose a novel multiple imputation-least absolute shrinkage and selection operator (MI-LASSO) variable selection method as an extension of the least absolute shrinkage and selection operator (LASSO) method to multiply-imputed data. The MI-LASSO method treats the estimated regression coefficients of the same variable across all imputed datasets as a group and applies the group LASSO penalty to yield a consistent variable selection across multiple-imputed datasets. We use a simulation study to demonstrate the advantage of the MI-LASSO method compared with the alternatives. We also apply the MI-LASSO method to the University of Michigan Dioxin Exposure Study to identify important circumstances and exposure factors that are associated with human serum dioxin concentration in Midland, Michigan.