Loss Of Predictive Power Research Articles

COMPARISON OF STATISTICAL AND MACHINE LEARNING APPROACHES IN A SURVEY FROM THE EPIDEMIOLOGY OF NON-COMMUNICABLE DISEASES

Objective: To determine if the disuse of basic machine learning approaches may cause a significant loss of predictive power in the epidemiology study. <Figure> Design and method: A cross-sectional study «Epidemiology of Cardiovascular Diseases in the Russian federation (ESSE-RF)» was performed in 13 regions in 2012–2013. ESSE-RF included randomly selected 21768 participants aged 25–64 with a response rate > 80 %. Standard epidemiology methods and criteria were used. The current sub-study included 13912 participants. We chose lasso regression (L1Regression) and random forest (RF) as basic machine learning methods and logistic regression (LogRegression) as a basic statistical method. We compared these algorithms on predicting arterial hypertension (AH) from 18 clinical, demographic, and social risk factors from international AH guidelines. We fitted the models on the training sample and assessed their quality on the holdout sample with an AUC metric. We used 1000 random train-holdout splits to exclude randomness. We studied the RF feature impacts with model-agnostic techniques: partial dependence plots (PDPs), feature importances, and feature interaction measures. Statistical analysis was performed using R 3.6.1. Results: LogRegression had better performance (82.11 ± 0.42) than L1Regression (81.90 ± 0.40) and RF (81.60 ± 0.43) respectively. Nine out of the top ten important variables in LogRegression were in the top ten important variables of RF, and vice versa. However, RF feature importance was more compatible with the risk guidelines than the LogRegression one (fig. 1A). In particular, LDL and HDL levels were insignificant in the LogRegression but significant in the RF. PDPs of RF were strictly monotonous and close to linear (fig. 1B). There were no features with an interaction measure greater than 35 % except for age. Using the previous analysis information, we constructed new interpretable features, which improved the AUC of basic LogRegression by 0.21. Conclusions: Data from the ESSE-RF study is homogeneous. Associations between features and AH are approximately linear and non-interactive. Therefore, it is correct to adhere to the basic interpretable statistical algorithms. However, machine learning methods can provide additional information that can improve understanding of risk factors influence.

Persistence and Performance in Co-Enrollment Network Embeddings: An Empirical Validation of Tinto's Student Integration Model

In this article, we empirically validate Tinto's Student Integration model, in particular, the predictions the model makes regarding both students' academic outcomes and their dropout decisions. In doing so, we analyze three decades' worth of student enrollments at an Australian university and present a novel methodological approach using graph embedding techniques to capture both structural and neighborhood-based features of the co-enrollment network. In keeping with Tinto's model, we find that not only do these embedded representations of students' social network predict their final grade point average (GPA), but also are able to successfully classify students who dropout. Our results show that these embedded representations of a student's social network can achieve F <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> -scores of up to 0.79 when classifying dropout and explain up to 10% of the variance in student's final GPA. When controlling for a small set of covariates and variables common to the literature, this performance increases to 0.83 and 24%, respectively. Furthermore, the performance of these methods is robust to both changes in their parameterization and to corruption of the underlying social networks. Importantly, this implies that hyperparameters may be selected to reduce the computational demands of this method without loss of predictive power. The novelty of this method, and its ability to identify student dropout, merits further investigation to preemptively identify at-risk students.

Performance Analysis of Learning Rate Parameter on Prediction of Signal Power Loss for Network Optimization and Better Generalization

This research work explores the neural network learning capabilities by using a multi-layer perceptron artificial neural network to predict signal power loss by means of dataset from long term evolution network. The analysis of the effect of the learning rate parameter and the adoption of early stopping method during network training have been executed by using varied values of learning rate to ascertain the best learning rate during the neural network training. Also, there were neural network training without the application of learning rate and early stopping method and comparison have been made with the output results as shown in different tables. Output results comparisons have been performed using training regression and performance mean squared error. Two back propagation training algorithms, the Levenberg–Marquardt and the Bayesian Regularization algorithms were employed for the network training and comparison of their prediction abilities examined using same values of learning rates and on application of early stopping method as well as without learning rate and without early stopping method. The result shows an optimal performance of the neural network model on application of 0.005 learning rate and using 75%:15%:15% early stopping method with training regression 0.99267 and performance mean squared error 2.47 using Levenberg–Marquardt and training regression 0.99488 and performance mean squared error of 1.910 using Bayesian Regularization algorithms, respectively. Without application of learning rate and early stopping method, training the network using Levenberg–Marquardt algorithm gives training regression of 0.97111 and performance mean squared error of 7.38 using Levenberg–Marquart algorithm and training regression of 0.99248 and performance mean squared error of 4.42 using Bayesian Regularization algorithm. The margin between the two output results demonstrates the impact and importance of learning rate parameter as well as adopting early stopping method for neural network training for network optimization and better network generalization.

Performance Comparison of Non-Linear Median Filter Built on MLP-ANN and Conventional MLP-ANN: Using Improved Dataset Training in Micro-Cell Environment

This research work explores the Levenberg- Marquardt training algorithm used for Artificial Neural Network (ANN) optimization during training and the Bayesian Regularization algorithm for the enhanced generalized trained network in training a designed non-linear vector median filter built on Multi-Layer Perceptron (MLP) ANN called model-1 and a conventional MLP ANN called model-2. The model-1 employed in the design helps in dataset de-noising to ensure the removal of unwanted signals for the improved training dataset. An early stopping method in the ratio of 80:10:10 for training, testing, and validation to overcome the problem of over-fitting during network training was employed. First-order statistical indices, the standard deviation, root mean squared error, mean absolute error, and correlation coefficient were adopted for network training analysis and comparative analysis of the designed model-1 and model-2, respectively. Two locations, Line-of-sight (location-1) and non-Line-of-Sight (location-2), were considered where the dataset was captured. The training results from the two locations for the two models demonstrated improved prediction of signal power loss using model-1 in comparison to model-2. For instance, the correlation coefficient, which shows the strength of the predicted value to the measured values (closer to 1) establishing a strong connection, gives 0.990 and 0.995 using model-1 for location-1, training with Lavenberg-Marquardt and Bayesian Regularization algorithm, respectively and 0.965 and 0.980 for model-2 using the same algorithms. It is seen that the Bayesian regularization algorithm, which optimizes the network in accordance with the Levenberg- Marquardt algorithm, gave better prediction results. The same sequence of improved perditions using designed model-1 in comparison to model-2 were seen with training results in location-2 while also adopting other employed 1st order statistical indices.

Enhanced Error Reduction of Signal Power Loss During Electromagnetic Propagation: Architectural Composition and Learning Rate Selection

This research work analyses the effect of the architectural composition of Multi-Layer Perceptron (MLP) Artificial Neural Network (ANN) combined with the effect of the learning rate for effective prediction of signal power loss during electromagnetic signal propagation. A single hidden layer and two hidden layers of MLP ANN have been considered. Different configurations of the neural network architecture ranging from 4 to 100 for both MLP networks have been analyzed. The required hidden layer neurons for optimal training of a single layer multi-layer network were 40 neurons with 0.99670 coefficient of correlation and 1.28020 standard deviations, while [68 72] trained two hidden layers multi-layer perceptron with 0.98880 coefficient of correlation and standard deviation of 1.42820. Different learning rates were also adopted for the network training. The results further validate better MLP neural network training for signal power loss prediction using single-layer perceptron network compared to two hidden layers perceptron network with the coefficient of correlation of 0.99670 for single-layer network and 0.9888 for two hidden layers network. Furthermore, the learning rate of 0.003 shows the best training capability with lower mean squared error and higher training regression compared to other values of learning rate used for both single layer and two hidden layers perceptron MLP networks.

Effect of Architectural Composition of MLP ANN in Neural Network Learning for Signal Power Loss Prediction

This work analyzes the architectural complexity of a Multi-Layer Perceptron (MLP) Artificial Neural Network (ANN) model suitable for modeling and predicting signal power loss in micro-cellular environments. The MLP neural network model with one, two, and three hidden layers respectively were trained using measurement datasets used as the target values collected from a micro-cell environment that is suitable to describe different propagation paths and conditions. The neural network training has been performed by applying different training techniques to ensure a well-trained network for good generalization and avoid over-fitting during network training. Bayesian regularization algorithm (that updates weights and biases during network training) following the Levenberg-Marquardt optimization training algorithm was used as the training algorithm. A comparative analysis of training results from one, two, and three hidden layers MLP neural networks show the best prediction result of the signal power loss using a neural network with one hidden layer. A complex architectural composition of the MLP neural network involved very high training time and higher prediction errors.

Comparison of machine learning approaches used to identify the drivers of Bakken oil well productivity

AbstractGeologists and petroleum engineers have struggled to identify the mechanisms that drive productivity in horizontal hydraulically fractured oil wells. The machine learning algorithms of Random Forest (RF), gradient boosting trees (GBT) and extreme gradient boosting (XGBoost) were applied to a dataset containing 7311 horizontal hydraulically fractured wells drilled into the middle member of the Bakken Formation from 2010 through 2017. The initial goal is to use these data‐driven machine learning algorithms to identify the most important explanatory predictors of well productivity within nine subareas and the composite area. Predictor variables representing initial gas production, the initial 180‐day water cut, and vertical depth vary spatially and are identified with geologically favorable areas. Well‐completion predictors include the well lateral length, number of fracture stages, volume of proppant per stage, and the volume of injected fluids per stage. The performance of methods is compared based on a common test sample. The analysis then examines the comparative predictive performance of the three algorithms for 1330 wells that had initiated production after the initial 7311 well sample had been producing. The computations of predictor importance identified the initial 180‐day water cut and the 30‐day initial gas production predictors as having a dominant influence in most subareas and for the composite area. The relative importance of well completion predictor variables, that is, the number of fracture stages per well, volume of injected proppant per stage, volume of injected fluids per stage, and lateral length, varied considerably across the subareas. For the common test or holdout sample, the models calibrated with the XGBoost algorithm had superior predictive power. The predictive power of all the algorithms trained on the data from the original sample suffered some loss when tested with a sample of wells that had started production after the end of that period. Implications of the empirical findings and strategies to mitigate loss of predictive power are discussed in the concluding section.

A Transient Mixed Elastohydrodynamic Lubrication Model for Helical Gear Contacts

Abstract In this study, a non-Newtonian, transient, isothermal, mixed elastohydrodynamic lubrication (EHL) model is proposed for helical gear contacts. The model accounts for nonelliptical contacts subject to spatially varying sliding and rolling velocity fields that are not aligned with any principal axis of the contact region, which is the case for helical gear contacts. The time-varying changes pertaining to key contact parameters and relative motion of roughness profiles on mating tooth surfaces are captured simultaneously to follow the contact from the root to the tip of a tooth while accounting for the transient effect due to relative motions of the roughness profiles. Actual tooth load distributions, contact kinematics, and compliances of helical gear contacts are provided to this model by an existing helical gear load distribution model. Measured three-dimensional roughness profiles covering the entire meshing zone are incorporated in the analyses to investigate its impact on the EHL conditions as well as mechanical power loss. Results of a parametric sensitivity study are presented to demonstrate the influence of operating conditions and surface roughness on the EHL behavior and the resultant gear mesh mechanical power loss of an example helical gear pair. The accuracy of the proposed mixed-EHL model is assessed by comparing the mechanical power loss predictions to available experimental results.

Frictional dynamic model predictions of FZG-A10 spur gear pairs considering profile errors

In this work, a nonlinear dynamic model of an FZG-A10 spur gear was investigated by taking into account for the actual time-varying gear mesh stiffness and the frictional effects between meshing gear teeth to evaluate the influence of the dynamic effects on frictional gear power loss predictions. The equations of motion of the generalized translational-torsional coupled dynamic system derived from Lagrange principle was extended compared to authors’ previous work in order to account for time dependent coefficient of friction and profile errors. The dynamic response of spur gears, computed by an iterative implicit scheme of Newmark, is changed due to the presence of coefficient of friction and profile errors. A dynamic analysis was performed and the influence of frictional effect including tooth shape deviations, in particular, was scrutinized since a time-dependent coefficient of friction is deeply related to the gear surface roughness and all parameters dependent on gears error profiles are introduced in the proposed model. The predicted meshing gear power losses with constant and local friction coefficient were compared. The influence of constant and variable profile errors considered in the local coefficient of friction formulation was also studied and their corresponding root mean square (RMS) power loss was compared to the experimental results. The results using FZG A10 spur gear pairs running under several operating conditions (different loads and speeds) validate the superiority of the proposed model against previous similar models.

Spatially resolved electrical modelling of cracks and other inhomogeneities in crystalline silicon solar cells

AbstractWe developed a fully scalable quasi‐SPICE approach for electrical modelling of stress‐induced inhomogeneities in crystalline silicon solar cells, based on damage detected by electroluminescence imaging. We present a model of active and inactive microcracks, highlight their effects on the current‐voltage characteristic, and validate the model on a number of experimental cases. The generalisation of our model to a complexly damaged solar cell shows excellent agreement with the measurements, with only 0.12% estimation error of the power loss of the complexly damaged cell. The ability to accurately model fundamental stress induced inhomogeneities, and their effects on cell level is crucial to estimating long‐term performance degradation and energy yield of installed photovoltaic modules throughout their lifetime. The developed scalable model paves a path to intricate, statistically supported modelling for power loss prediction and mitigation in solar cells, photovoltaic modules, and photovoltaic arrays caused by mechanically induced inhomogeneities.

Parameter Identification of a Power Loss Model for Vehicle Transmissions Based on Sensitivity Analysis

&lt;div class="section abstract"&gt;&lt;div class="htmlview paragraph"&gt;As the transmission design directly impacts drive unite operation and power flow to the driveline, the transmission power loss is a critical target in the drivetrain development. The demand of more precise and more efficient power loss prediction has therefore increased significantly, which highlights the need of new methodologies in order to optimize the power loss model for vehicle transmissions. The possible power losses that exist in the power flow path, are gear mesh losses, gear churning losses, gear windage losses, bearing losses, synchronizer losses and sealing losses. Thanks to the decades of research, analytical models are available for the prediction of these component losses, which could deliver power loss distributions and overall efficiency maps of complex transmissions.&lt;/div&gt;&lt;div class="htmlview paragraph"&gt;The aim of this paper is to introduce a methodology to improve the accuracy of a chosen power loss model on a system level. A detailed power loss prediction for a two-speed transmission in an electric vehicle has been performed. The simulated overall power losses and the available experimental results match well. However, since many assumptions are made in the analytical modelling process, there are still deviations between the predicted and the measured results. In order to reduce the deviations, all uncertain parameters are firstly analyzed based on the parameter sensitivity analysis method FAST that allows determining the influential uncertain parameters. The sensitivities of those influential parameters are locally defined at all operating regions, by which the sensitive operating areas of all influential uncertain parameters to the overall power losses could be defined. The identification of these parameters at their sensitive regions prevents the unnecessary interference with other uncertain parameters at the identification process. With the help of identified parameters, a better proximity between the simulated and measured results is achieved.&lt;/div&gt;&lt;/div&gt;

A Power Loss Measurement Technique for Inductive Power Transfer Magnetic Couplers

A critical aspect of designing an inductive power transfer system is the accurate characterization and prediction of power losses in its magnetic couplers under all operating conditions. Understanding the loss mechanisms and their distribution can help the optimization of the magnetic couplers while also being essential to determining thermal management requirements and optimal control parameters. Previously proposed loss measurement methods are neither able to accurately predict the losses in couplers nor characterize their distribution. As a solution, this article proposes a new experimental method, termed the stepped resonant excitation (SRE) method that can accurately predict the power losses in magnetic couplers operated under practical conditions. In addition, the SRE method is capable of producing better insight to the loss distribution by separately quantifying the winding loss and core loss. Experimental results presented in this article show that the SRE method can predict the core loss of an accurately characterized toroidal inductor with less than 10% error when excited with magnetic fields higher than 50 mT. Additionally, the losses measured using the SRE method in a double-D magnetic coupler match well with finite element analysis simulations conducted using datasheet loss curves at 25 °C.

Oil distribution and churning losses of gearboxes: Experimental and numerical analysis

Currently, energy efficiency represents one of the main requisites in mechanical design. Gearboxes play an important role in several industrial applications and, even if they are already characterized by high efficiency, their performance can still be improved, leading to energy saving and the reduction of pollutant emissions. The possibility to investigate different gearbox geometries and operating conditions in the preliminary stages of design represents a clear advantage for engineers and can significantly contribute to the achievement of these objectives. However, analytical formulations that are able to accurately describe gear losses, especially those due to the interaction with the lubricant, are not available or are not accurate enough. Furthermore, the analytical models are typically not suitable for novel gearbox designs.In the present work, a numerical approach based on Computational Fluid Dynamics (CFD) for the prediction of lubricant fluxes and power losses of gearboxes is presented. The presented methodology utilizes a meshing strategy that reduces the computational effort of the simulations and enables parametrical studies. The numerical approach is validated both with torque measurements and data regarding the oil flow obtained by the Particle Image Velocimetry (PIV). Thus, the validation of the numerical approach is performed not only qualitatively with respect to the oil distribution but also quantitatively with respect to the torque and velocity field in the oil sump. Good agreement of the CFD and experimental data was observed for the torque, oil distribution and fine flow details including the vorticial structures and recirculation areas.

Multistage Collaborative Machine Learning and its Application to Antenna Modeling and Optimization

A multistage collaborative machine learning (MS-CoML) method that can be applied to efficient multiobjective antenna modeling and optimization is proposed. Machine learning methods, including single-output Gaussian process regression (SOGPR) and symmetric and asymmetric multioutput GPR (MOGPR) methods, are introduced to collaboratively build highly accurate multitask surrogate models for antennas. Variable-fidelity electromagnetic (EM) models are simulated, with their responses utilized to build separate MOGPR surrogate models. By combining the three machine-learning methods in a multistage framework, mappings between the same and different responses of the EM models with variable fidelity are learned, therein helping to substantially reduce the computational effort under a negligible loss of predictive power. Three antenna designs aiming at single-band, broadband, and multiband applications are selected as examples. And, for illustrating the applicability and superiority of the proposed MS-CoML method, a reference point-based multiobjective antenna optimization algorithm is used to optimize these three antennas. Simulation results show that using the MS-CoML method can significantly reduce the total optimization time without compromising modeling accuracy and optimized performance.

Innovative Meshing Strategies for Bearing Lubrication Simulations

Efficiency improvement is the new challenge in all fields of design. In this scenario the reduction of power losses is becoming more and more a main concern also in the design of power transmissions. Appropriate models to predict power losses are therefore required from the earliest stages of the design phase. The aim of this project is to carry out lubrication simulations of several variants of a cylindrical roller bearing to understand the lubricant distribution and the related churning power losses. Several strategies to reduce the computational effort were used. Among them the sectorial symmetry and three innovative meshing strategies (purely analytical with and without interfaces and analytical/subtractive) that were implemented in the OpenFOAM® environment. The results of the different approaches were compared among them and reasonable savings in terms of computational effort were shown.

Design of Novel Gas Foil Thrust Bearings and Test Validation in a High-Speed Test Rig

Abstract Small gas foil bearings (FBs) with shaft diameter below 25 mm can find many applications in air compressors for fuel cells, electrical turbo chargers, small unmanned air vehicles, turbo alternators, etc. These small machines are characterized by very light load to the radial FBs, and thus rotordynamics stability is more challenging than load capacity. However, a main challenge of gas foil thrust bearings (GFTBs) is how to increase the load capacity, and the challenge remains the same regardless of the size. In previous publications on experimental studies on GFTBs, the measured load capacity is well below the prediction due to challenges in testing as well as manufacturing of GFTBs. Difficulty in achieving the design load capacity often leads to increasing the bearing size in actual applications with penalty of higher power loss. This paper presents design feature of a novel GFTB with outer diameter of 38 mm and static performance up to 155 krpm under external load of 75 N using a high-speed test rig. The 38 mm GFTB presented in this paper is a three-layered structure for easy design and manufacturing, and the unique design feature allows easy scale down and scale up to different sizes. Reynolds equations for compressible gas and the two-dimensional thin plate model were adopted for fluid–structure interaction simulation to predict load capacity and power loss of the GFTB. The predicted power loss and load capacity agree well with the measurements.

Power Loss Prediction for Distributed Energy Resources: Rapid Loss Estimation Equation

The rapid expansion of distributed energy resources has led to increasingly complex systems with numerous power converters. Accurate converter loss prediction in large grids and microgrids is essential for financial and reliability evaluation. Existing system-level analysis focuses on distribution losses and oversimplifies converter losses by assuming fixed efficiency. However, converter losses are highly variable under different operating conditions. Moreover, commercially-available multidomain simulation tools are too slow to be applied to system-level analysis. To provide computationally simple loss prediction under all operating conditions, the rapid loss estimation (RLE) equation is proposed. First, the real operating conditions of the converter are determined for the intended application. Then, accurate loss information is extracted from detailed converter behavior in multidomain simulations. Finally, the RLE equation is obtained: a parametric equation, which is fast enough for system-level simulation while capturing the converter's complexity at different operating conditions. A dc microgrid with three different converters, one each for solar generation, electric vehicle charging stations and battery storage, is considered to highlight the benefits of the proposed loss estimation tool.

Predicting LGD distributions with mixed continuous and discrete ordinal outcomes

The loss given default (LGD) distribution is known to have a complex structure. Consequently, the parametric approach for its prediction by fitting a density function may suffer a loss of predictive power. To overcome this potential drawback, we use the cumulative probability model (CPM) to predict the LGD distribution. The CPM applies a transformed variable to model the LGD distribution. This transformed variable has a semiparametric structure. It models the predictor effects parametrically. The functional form of the transformation is unspecified. Thus, CPM provides more flexibility and simplicity in modeling the LGD distribution. To implement CPM, we collect a sample of defaulted debts from Moody’s Default and Recovery Database. Given this sample, we use an expanding rolling window approach to investigate the out-of-time performance of CPM and its alternatives. Our results confirm that CPM is better than its alternatives, in the sense of yielding more accurate LGD distribution predictions.

Interpretation of Urban Power Consumers Behaviors to Predict Power Loss in summer

High Temperature In The Summer Of India. Interpretation Of Electricity Consumption Is Crucial In Summer For Urban Consumers. We Are Focus Here For Only Indian Summer Urban Customers Energy Consumption To Analysis And Predict Behavior Of Electricity Theft. Data Mining Techniques Are Employing To Analyses Indian Summer Urban Customers. Online Sequential Machine And Support Vector Machine Is Used For This Behaviors Classification And Prediction.Mainly we focus Support vector machine to classified consumers and online sequential machine is used to detect and predict consumers behaviors.

Influence of crate size, oviposition time, number of adults and cannibalism on the reproduction ofTenebrio molitor

In an industrialised mealworm farm it is important to maximise the production and to know the number of mealworms in each container as early as possible in a fast and reliable way. Two experiments were performed. The first experiment assessed the influence of the number of beetles, crate surface area and oviposition time on the number produced mealworms. A full factorial design was used with 11 beetle densities (between 2.3-300 mg beetles/cm2), 5 oviposition times (from 1-14 days) and 4 different crate sizes (between 250-2,000 cm2). In the second experiment, the influence of cannibalism on the number of produced mealworms was assessed via an alternative oviposition method that prohibited cannibalism. Multiple linear regression was used to model the results. The results indicate that the number of beetles, oviposition time and surface area could predict the number of produced mealworms well. An increase in one of the three parameters increased the number of produced mealworms without reaching an optimum. Furthermore, the number of beetles and the oviposition time can be combined in one parameter, beetledays with minimal loss of predictive power of the model. Nevertheless, the number of produced mealworms per female did decline rapidly with increasing oviposition time and density. The latter is, at least in part, due to cannibalism, as the second experiment indicates that the density effect is almost eliminated when the beetles are unable to reach their eggs. In conclusion, this study indicates that it is possible to construct a formula that can be used toa priori determine the final number of produced mealworms based on the number of beetles, surface area and oviposition time and that cannibalism can greatly reduce the number of produced mealworms. Reducing cannibalism can greatly increase the efficiency and therefore production of a mealworm farm.