High Level Of Prediction Accuracy Research Articles

Predicting Diabetic Neuropathy Risk Level Using Artificial Neural Network and Clinical Parameters of Subjects With Diabetes.

Background:A risk assessment tool has been developed for automated estimation of level of neuropathy based on the clinical characteristics of patients. The smart tool is based on risk factors for diabetic neuropathy, which utilizes vibration perception threshold (VPT) and a set of clinical variables as potential predictors.Methods:Significant risk factors included age, height, weight, urine albumin-to-creatinine ratio, glycated hemoglobin, total cholesterol, and duration of diabetes. The continuous-scale VPT was recorded using a neurothesiometer and classified into three categories based on the clinical thresholds in volts (V): low risk (0-20.99 V), medium risk (21-30.99 V), and high risk (≥31 V).Results:The initial study had shown that by just using patient data (n = 5088) an accuracy of 54% was achievable. Having established the effectiveness of the “classical” method, a special Neural Network based on a Proportional Odds Model was developed, which provided the highest level of prediction accuracy (>70%) using the simulated patient data (n = 4158).Conclusion:In the absence of any assessment devices or trained personnel, it is possible to establish with reasonable accuracy a diagnosis of diabetic neuropathy by means of the clinical parameters of the patient alone.

Chaotic Time Series Prediction Using LSTM with CEEMDAN

Chaotic systems are complex dynamical systems that play a very important role in the study of the atmosphere, aerospace engineering, finance, etc. To improve the accuracy of chaotic time series prediction, this study proposes a hybrid model CEEMDAN-LSTM which combines Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) and long short-term memory (LSTM). In the model, the original time series is decomposed into several intrinsic mode functions (IMFs) and a residual component. To reduce the difficulty of predicting chaotic time series and provide a high level of predictive accuracy, the LSTM prediction model is built for all each characteristic series from CEEMDAN deposition. Finally, the final prediction results are obtained by combining all the prediction sequences. To test the effectiveness of this model we proposed, we examined the CEEMDAN-LSTM model using the Lorenz-63 system. Further compared to Autoregressive Integrated Moving Average (ARIMA ), Support Vector Regression (SVR), multilayer perceptron (MLP), and the single LSTM model, the results of the experiment show that the proposed model performs better in the prediction of chaotic time series. Besides, the hybrid model proposed in this paper has better results than the LSTM model alone. Therefore, hybrid models based on deep learning methods and signal decomposition methods have great potential in the field of chaotic time series prediction.

Predicting Alert Source Device using Machine Learning Algorithms

In a large distributed virtualized environment, predicting the alerting source from its text seems to be daunting task. This paper explores the option of using machine learning algorithm to solve this problem. Unfortunately, our training dataset is highly imbalanced. Where 96% of alerting data is reported by 24% of alerting sources. This is the expected dataset in any live distributed virtualized environment, where new version of device will have relatively less alert compared to older devices. Any classification effort with such imbalanced dataset present different set of challenges compared to binary classification. This type of skewed data distribution makes conventional machine learning less effective, especially while predicting the minority device type alerts. Our challenge is to build a robust model which can cope with this imbalanced dataset and achieves relative high level of prediction accuracy. This research work stared with traditional regression and classification algorithms using bag of words model. Then word2vec and doc2vec models are used to represent the words in vector formats, which preserve the sematic meaning of the sentence. With this alerting text with similar message will have same vector form representation. This vectorized alerting text is used with Logistic Regression for model building. This yields better accuracy, but the model is relatively complex and demand more computational resources. Finally, simple neural network is used for this multi-class text classification problem domain by using keras and tensorflow libraries. A simple two layered neural network yielded 99 % accuracy, even though our training dataset was not balanced. This paper goes through the qualitative evaluation of the different machine learning algorithms and their respective result. Finally, two layered deep learning algorithms is selected as final solution, since it takes relatively less resource and time with better accuracy values.

Development of prediction model through linear multiple regression for the prediction and analysis of the GSM of embroidered fabric

PurposeThe purpose of this paper is to establish the regression equation based upon a set of samples prepared through structured design of experiment and form a prediction model for prediction of the areal density gram per square meter (GSM) of the embroidered fabrics and study the influence of basic input parameters.Design/methodology/approachEmbroidery samples are prepared taking input parameters as GSM of the base fabric, linear density of the embroidery thread and stitch density of the embroidery design. Three levels of values are identified for each of the input parameters. Taguchi and Box-Behnken experiment design principles are used to prepare two sets of samples. Linear multiple regression is used to determine the prediction equations based upon each of the two sets and the combined set as well. Prediction equations are statistically verified for the prediction accuracy. Also, surface curves are prepared to study the influence of embroidery parameters on the GSM.FindingsIt is found that all the three prediction models developed in this study can predict with a very satisfactory level of accuracy. However, the regression equation based upon the data set prepared according to Taguchi experiment design is emerged as the prediction model with highest level of prediction accuracy. Corresponding equation coefficients and several three-dimensional surface curves are used to study the influence of embroidery parameters and it is found that the stitch density is the most influential input parameter followed by stitch length and the GSM of base fabric.Research limitations/implicationsThis can be used to assess the GSM of embroidered fabrics before starting the actual embroidery process. So, this model can help the embroidery designers significantly to pre-estimate the GSM of the embroidered fabrics and select the design parameters accordingly. Also, this model can be a useful tool for estimation of thread consumption and thread cost in embroidery.Practical implicationsThe input parameters used here are very basic parameters related to design and materials, which can be easily available. And also, a simple linear multiple regression is used to make the prediction equation simple and easy to use. So, this model can help the embroidery designers or garment designers to select/adjust the embroidery parameters and thread parameters accordingly in the planning and designing stage itself to ensure that the GSM of embroidered fabrics remains within desirable range. Also, this prediction model developed hereby may be a very useful tool for estimation of the consumption and cost of embroidery threads.Originality/valueThis paper presents a very fundamental study to reveal the effect of embroidery parameters on the GSM, through development of regression equations. It can help future researchers in optimizations of input parameters and forming a technical guideline for the embroidery designers for selection of the design parameters for a desired GSM of embroidered fabric.

Predicting Poverty Using Geospatial Data in Thailand

Poverty statistics are conventionally compiled using data from household income and expenditure survey or living standards survey. This study examines an alternative approach in estimating poverty by investigating whether readily available geospatial data can accurately predict the spatial distribution of poverty in Thailand. In particular, geospatial data examined in this study include night light intensity, land cover, vegetation index, land surface temperature, built-up areas, and points of interest. The study also compares the predictive performance of various econometric and machine learning methods such as generalized least squares, neural network, random forest, and support vector regression. Results suggest that intensity of night lights and other variables that approximate population density are highly associated with the proportion of an area’s population who are living in poverty. The random forest technique yielded the highest level of prediction accuracy among the methods considered in this study, perhaps due to its capability to fit complex association structures even with small and medium-sized datasets. Moving forward, additional studies are needed to investigate whether the relationships observed here remain stable over time, and therefore, may be used to approximate the prevalence of poverty for years when household surveys on income and expenditures are not conducted, but data on geospatial correlates of poverty are available.

Multiple steps ahead solar photovoltaic power forecasting based on univariate machine learning models and data re-sampling

Abstract Accurate predictions of solar Photovoltaic (PV) power generation at different time scales are essential for reliable operations of energy management systems. Contemporary methods can accurately predict solar PV power a few steps ahead but fail to maintain high level of prediction accuracy as the number of steps increases. In this paper, we present a simple but effective univariate approach to predict solar PV power output multiple steps ahead. The objective is to maintain high level accuracy of the machine learning algorithms for multiple steps ahead prediction. The novelty of the proposed approach lies in the integration of a data re-sampling technique with machine learning algorithms. The data re-sampling technique enables machine learning algorithms to compute multiple steps ahead predictions via simple single-step ahead prediction on the re-sampled time-series. At each prediction step, the proposed approach first obtains a new representation of the original time series based on the re-sampling process. A separate prediction model is then implemented for single-step prediction on the re-sampled time series. The single-step ahead prediction of re-sampled time series corresponds to m-th step ahead prediction of original time series as determined by the re-sampling factor m. Since the proposed approach computes multiple steps ahead predictions by combining a set of models designed for single-step ahead prediction, it significantly reduces the error at higher prediction steps. We have evaluated the effectiveness of the proposed approach for 5-min to 3-h ahead predictions using 2 years of data from a 1.22 MW PV system in Australia.

Machine Learning Models for Analysis of Vital Signs Dynamics: A Case for Sepsis Onset Prediction.

Objective Achieving accurate prediction of sepsis detection moment based on bedside monitor data in the intensive care unit (ICU). A good clinical outcome is more probable when onset is suspected and treated on time, thus early insight of sepsis onset may save lives and reduce costs. Methodology We present a novel approach for feature extraction, which focuses on the hypothesis that unstable patients are more prone to develop sepsis during ICU stay. These features are used in machine learning algorithms to provide a prediction of a patient's likelihood to develop sepsis during ICU stay, hours before it is diagnosed. Results Five machine learning algorithms were implemented using R software packages. The algorithms were trained and tested with a set of 4 features which represent the variability in vital signs. These algorithms aimed to calculate a patient's probability to become septic within the next 4 hours, based on recordings from the last 8 hours. The best area under the curve (AUC) was achieved with Support Vector Machine (SVM) with radial basis function, which was 88.38%. Conclusions The high level of predictive accuracy along with the simplicity and availability of input variables present great potential if applied in ICUs. Variability of a patient's vital signs proves to be a good indicator of one's chance to become septic during ICU stay.

TransPrise: a novel machine learning approach for eukaryotic promoter prediction.

As interest in genetic resequencing increases, so does the need for effective mathematical, computational, and statistical approaches. One of the difficult problems in genome annotation is determination of precise positions of transcription start sites. In this paper we present TransPrise—an efficient deep learning tool for prediction of positions of eukaryotic transcription start sites. Our pipeline consists of two parts: the binary classifier operates the first, and if a sequence is classified as TSS-containing the regression step follows, where the precise location of TSS is being identified. TransPrise offers significant improvement over existing promoter-prediction methods. To illustrate this, we compared predictions of TransPrise classification and regression models with the TSSPlant approach for the well annotated genome of Oryza sativa. Using a computer equipped with a graphics processing unit, the run time of TransPrise is 250 minutes on a genome of 374 Mb long. The Matthews correlation coefficient value for TransPrise is 0.79, more than two times larger than the 0.31 for TSSPlant classification models. This represents a high level of prediction accuracy. Additionally, the mean absolute error for the regression model is 29.19 nt, allowing for accurate prediction of TSS location. TransPrise was also tested in Homo sapiens, where mean absolute error of the regression model was 47.986 nt. We provide the full basis for the comparison and encourage users to freely access a set of our computational tools to facilitate and streamline their own analyses. The ready-to-use Docker image with all necessary packages, models, code as well as the source code of the TransPrise algorithm are available at (http://compubioverne.group/). The source code is ready to use and customizable to predict TSS in any eukaryotic organism.

Predicting and explaining corruption across countries: A machine learning approach

In the era of Big Data, Analytics, and Data Science, corruption is still ubiquitous and is perceived as one of the major challenges of modern societies. A large body of academic studies has attempted to identify and explain the potential causes and consequences of corruption, at varying levels of granularity, mostly through theoretical lenses by using correlations and regression-based statistical analyses. The present study approaches the phenomenon from the predictive analytics perspective by employing contemporary machine learning techniques to discover the most important corruption perception predictors based on enriched/enhanced nonlinear models with a high level of predictive accuracy. Specifically, within the multiclass classification modeling setting that is employed herein, the Random Forest (an ensemble-type machine learning algorithm) is found to be the most accurate prediction/classification model, followed by Support Vector Machines and Artificial Neural Networks. From the practical standpoint, the enhanced predictive power of machine learning algorithms coupled with a multi-source database revealed the most relevant corruption-related information, contributing to the related body of knowledge, generating actionable insights for administrator, scholars, citizens, and politicians. The variable importance results indicated that government integrity, property rights, judicial effectiveness, and education index are the most influential factors in defining the corruption level of significance.

EXTREME GRADIENT BOOSTING METHOD IN THE PREDICTION OF COMPANY BANKRUPTCY

Abstract Machine learning methods are increasingly being used to predict company bankruptcy. Comparative studies carried out on selected methods to determine their suitability for predicting company bankruptcy have demonstrated high levels of prediction accuracy for the extreme gradient boosting method in this area. This method is resistant to outliers and relieves the researcher from the burden of having to provide missing data. The aim of this study is to assess how the elimination of outliers from data sets affects the accuracy of the extreme gradient boosting method in predicting company bankruptcy. The added value of this study is demonstrated by the application of the extreme gradient boosting method in bankruptcy prediction based on data free from the outliers reported for companies which continue to operate as a going concern. The research was conducted using 64 financial ratios for the companies operating in the industrial processing sector in Poland. The research results indicate that it is possible to increase the detection rate for bankrupt companies by eliminating the outliers reported for companies which continue to operate as a going concern from data sets.

Identification of Cigarette Brands by Soft Independent Modeling of Class Analogy of Volatile Substances.

IntroductionThis study aimed to develop a method for discriminating cigarette brands based on the profiles of volatile components extracted from the tobacco fraction of the finished cigarettes to authenticate branded cigarettes of unknown origin.MethodsAn analytical method comprising direct thermal desorption coupled with gas chromatography-quadrupole time-of-flight mass spectrometry was developed for acquiring volatile profiles of cigarettes. About 290 samples of commercially available cigarettes were analyzed. Within this batch, 123 samples represented four popular cigarette brands. They were selected for in-depth characterization. Multivariate analysis was used to investigate the interrelations among volatile compounds of cigarettes and to identify characteristic markers for the cigarette discrimination. Supervised pattern recognition techniques were used for designing classification models.ResultsPrincipal component analysis covering all detected volatiles allowed the differentiation of cigarettes based on the brand. A number of 56 volatile components were identified as markers with high discrimination power. These compounds were used for establishing classification models. A method of soft independent modeling of class analogy developed for the four studied cigarette brands proved to be efficient in the classification of unknown cigarettes, with accuracy between 95.9% and 100%.ConclusionsThe data evaluation by soft independent modeling of class analogy was highly accurate in classification of unknown cigarettes with a low rate of false positives and false negatives. The developed models can be used for discrimination of genuine from non-genuine products with high level of probability.ImplicationsProfiling of volatiles, which is commonly used for authentication of different food commodities, was applied for the characterization of cigarette tobacco for the purpose of authentication a cigarette brand. Volatile components with a high discrimination power were identified by means of multivariate statistical methods and used for establishing of a classification model. The classification model was able to discriminate genuine from non-genuine cigarettes with a high level of prediction accuracy. This model could be a powerful tool for tobacco control to judge the authenticity of cigarettes.

A library for wall-modelled large-eddy simulation based on OpenFOAM technology

This work presents a feature-rich open-source library for wall-modelled large-eddy simulation (WMLES), which is a turbulence modelling approach that reduces the computational cost of standard (wall-resolved) LES by introducing special treatment of the inner region of turbulent boundary layers (TBLs). The library is based on OpenFOAM and enhances the general-purpose LES solvers provided by this software with state-of-the-art wall modelling capability. The included wall models belong to the class of wall-stress models that account for the under-resolved turbulent structures by predicting and enforcing the correct local value of the wall shear stress. A review of this approach is given, followed by a detailed description of the library, discussing its functionality and extensible design. The included wall-stress models are presented, based on both algebraic and ordinary differential equations. To demonstrate the capabilities of the library, it was used for WMLES of turbulent channel flow and the flow over a backward-facing step (BFS). For each flow, a systematic simulation campaign was performed, in order to find a combination of numerical schemes, grid resolution and wall model type that would yield a good predictive accuracy for both the mean velocity field in the outer layer of the TBLs and the mean wall shear stress. The best result, ≈1% error in the above quantities, was achieved for channel flow using a mildly dissipative second-order accurate scheme for the convective fluxes applied on an isotropic grid with 27000 cells per δ3-cube, where δ is the channel half-height. In the case of flow over a BFS, this combination led to the best agreement with experimental data. An algebraic model based on Spalding’s law of the wall was found to perform well for both flows. On the other hand, the tested more complicated models, which incorporate the pressure gradient in the wall shear stress prediction, led to less accurate results. Program SummaryProgram Title: libWallModelledLESProgram Files doi:http://dx.doi.org/10.17632/m8dnsnp4nd.1Licensing provisions: GPLv3Programming language: C++Nature of problem: Large-eddy simulation (LES) is a scale-resolving turbulence modelling approach providing a high level of predictive accuracy. However, LES of high Reynolds number wall-bounded flows is prohibitively computationally expensive due to the need for resolving the inner region of turbulent boundary layers (TBLs) [1]. This inhibits the application of LES to many industrially relevant flows [2] and prompts for the development of novel modelling techniques that would modify the LES approach in a way that allows it to retain its accuracy (at least away from walls) yet significantly lower its computational cost. Solution method: Wall-modelled LES (WMLES) is an approach that is based on complementing LES with special near-wall modelling that allows to leave the inner layer of TBLs unresolved by the computational grid. Many types of wall models have been proposed [1,3], commonly tested within the framework of in-house research codes. Here, an open-source library implementing several wall models is presented. The library is based on OpenFOAM, which is currently the most widely used general-purpose open-source software for computational fluid dynamics. The developed library can be directly applied to both academic and industrial flow cases, leading to a wider adoption of wall modelling and better understanding of its strengths and limitations.[1] J. Larsson, S. Kawai, J. Bodart, and I. Bermejo-Moreno. Large eddy simulation with modeled wall-stress: recent progress and future directions. Mechanical Engineering Reviews, 3(1):1-23, 2016.[2] J. Slotnick, A. Khodadoust, J. Alonso, D. Darmofal, W. Gropp, E. Lurie, D. Mavriplis. CFD vision 2030 study: A path to revolutionary computational aerosciences, Tech. rep., NASA, 2014.[3] S. T. Bose and G. I. Park. Wall-modeled large-eddy simulation for complex turbulent flows. Annual Review of Fluid Mechanics, 50(1):535–561, 2018.

Text mining as a facilitating tool for deploying blockchain technology in the intellectual property rights system

The aim of the study is to introduce a new application of machine-learning techniques (text mining, clustering and classification) and the blockchain technology within the intellectual property rights (IPRs) management system. Using such machine-learning techniques facilitates the management process of intellectual properties (IPs) and makes it more efficient. Additionally, using the blockchain technology for IPRs management purposes enables all stakeholders to utilise the extracted data of the IP objects from the blockchain network. In this study, a text-mining technique was used to identify the two types of IP documents based on specific categories, namely, patent and trademark. In order to achieve this objective, a range of machine-learning techniques was used for 5,500 patent documents and 400 trademark documents. The results of the logistic regression model showed a high level of prediction accuracy of document type at the pre-registration stage on the blockchain network. This high level of prediction accuracy demonstrates that using machine-learning and text-mining techniques will facilitate the IPRs management system. This new application of specific machine-learning techniques in the IPRs management process contributes essentially to solving the problem in a conventional IPRs system associated with rights protection and data availability.

Predicting Currency Crises: A Novel Approach Combining Random Forests and Wavelet Transform

We propose a novel approach that combines random forests and the wavelet transform to model the prediction of currency crises. Our classification model of random forests, built using both standard predictors and wavelet predictors, and obtained from the wavelet transform, achieves a demonstrably high level of predictive accuracy. We also use variable importance measures to find that wavelet predictors are key predictors of crises. In particular, we find that real exchange rate appreciation and overvaluation, which are measured over a horizon of 16–32 months, are the most important.

Predicting ascites incidence in a simulated altitude-challenge using single nucleotide polymorphisms identified in multi-generational genome wide association studies.

Assessing pedigreed broiler lines for ascites resistance in an industry setting is time consuming. Further, the use of sibling selection implies study subjects are not used in the breeding program, and instead, siblings take their place in pedigree systems, which reduces overall genetic accuracy. The purpose of this study is to evaluate the effectiveness of prediction models produced with SNP with the goal of predicting ascites incidence. Ascites is the manifestation of a series of adverse changes in a broiler beginning with hypoxia. Increased blood pressure, accumulation of fluid in the abdominal cavity, and death can result. Ascites results in losses estimated at $100 million/year in the USA. A multi-generational genome wide association study in an unselected line maintained at the University of Arkansas since the 1990s identified chromosomal regions associated with ascites incidence in males when challenged at high altitude. From the identified regions of significance 20 SNP were selected to construct a predictive model (8 SNP on chromosome 11, and 12 SNP on chromosome Z). Ascites phenotype and genotype data were obtained for 295 male and female individuals from the REL line. Five modeling techniques were compared for their ascites predictive ability using a 70/30 split between training and validation. For both males and females, the artificial neural network model was the best fit prediction model due to the large area under the curve value of 0.997 and 0.997, respectively, as well as a low misclassification ratio of 0.027 and 0.037, respectively. Using a parameter decreasing method, the total number of SNP inputs used to construct artificial neural network (ANN) models was reduced. A 13 SNP male ANN model and an 18 SNP female ANN model were constructed with equally high levels of prediction accuracy compared with the 20 SNP input models. The construction of predictive ANN models indicates that we have found the genetic predictors to ascites outcome in male and female broilers from an elite line of the 1990s with a high level of accuracy.

Industry characteristic of bankruptcy prediction models appliance

The aim of the research is to develop the methodology of bankruptcy prediction applying the specified statutory values of the existing models with a glance to company’s industry and developing the author’s prediction model. Initially authors estimated the forecast accuracy of the existing models for the enterprises of 8 industries. Using CART (Classification And Regression Tree) methodology the original statutory values of the models were specified for every industry under research. The calculated statutory values demonstrated the high level of prediction accuracy and balanced the indicators of accuracy for bankrupt and non-bankrupt companies. The indicators with the maximum level of significance for bankruptcy prediction were selected from all the models. They formed a basis for a new developed model, which has demonstrated the high level of prediction accuracy on a sample under research. The statutory values for the new model were also developed.The implementation of the research’s results will increase the efficiency of bankruptcy prediction and low the number of bankrupt companies.

Towards an Online Seizure Advisory System-An Adaptive Seizure Prediction Framework Using Active Learning Heuristics.

In the last decade, seizure prediction systems have gained a lot of attention because of their enormous potential to largely improve the quality-of-life of the epileptic patients. The accuracy of the prediction algorithms to detect seizure in real-world applications is largely limited because the brain signals are inherently uncertain and affected by various factors, such as environment, age, drug intake, etc., in addition to the internal artefacts that occur during the process of recording the brain signals. To deal with such ambiguity, researchers transitionally use active learning, which selects the ambiguous data to be annotated by an expert and updates the classification model dynamically. However, selecting the particular data from a pool of large ambiguous datasets to be labelled by an expert is still a challenging problem. In this paper, we propose an active learning-based prediction framework that aims to improve the accuracy of the prediction with a minimum number of labelled data. The core technique of our framework is employing the Bernoulli-Gaussian Mixture model (BGMM) to determine the feature samples that have the most ambiguity to be annotated by an expert. By doing so, our approach facilitates expert intervention as well as increasing medical reliability. We evaluate seven different classifiers in terms of the classification time and memory required. An active learning framework built on top of the best performing classifier is evaluated in terms of required annotation effort to achieve a high level of prediction accuracy. The results show that our approach can achieve the same accuracy as a Support Vector Machine (SVM) classifier using only of the labelled data and also improve the prediction accuracy even under the noisy condition.

Quantitative Estimation of Plasma Free Drug Fraction in Patients With Varying Degrees of Hepatic Impairment: A Methodological Evaluation

Quantitative prediction of unbound drug fraction (fu) is essential for scaling pharmacokinetics through physiologically based approaches. However, few attempts have been made to evaluate the projection of fu values under pathological conditions. The primary objective of this study was to predict fu values (n = 105) of 56 compounds with or without the information of predominant binding protein in patients with varying degrees of hepatic insufficiency by accounting for quantitative changes in molar concentrations of either the major binding protein or albumin plus alpha 1-acid glycoprotein associated with differing levels of hepatic dysfunction. For the purpose of scaling, data pertaining to albumin and α1-acid glycoprotein levels in response to differing degrees of hepatic impairment were systematically collected from 919 adult donors. The results of the present study demonstrate for the first time the feasibility of physiologically based scaling fu in hepatic dysfunction after verifying with experimentally measured data of a wide variety of compounds from individuals with varying degrees of hepatic insufficiency. Furthermore, the high level of predictive accuracy indicates that the inter-relation between the severity of hepatic impairment and these plasma protein levels are physiologically accurate. The present study enhances the confidence in predicting fu in hepatic insufficiency, particularly for albumin-bound drugs.

Prediction of Soil-Water Characteristic Curve for Unbound Material Using Fredlund–Xing Equation-Based ANN Approach

AbstractMost of the existing soil-water characteristic curve (SWCC) prediction models do not have a high level of prediction accuracy. The R2 values of these model predictions range from 0.1 to 0.6...

An optimised logarithmic discretisation approach for accurate and efficient compact thermal models

The accuracy of building energy simulations is of considerable interest as discrepant results can elicit adverse financial and environment consequences. The physical and temporal scales considered within building energy applications necessitate compact modelling approaches. The prediction accuracy of such simulations is intrinsically linked with the ability to predict the thermal responses of structural elements. The optimal means of representing these components such that accurate solutions are ensured at minimal computational cost remains unclear. The current study seeks to optimise the spatial placement of nodes through assessing and reporting results pertaining to a logarithmic spatial discretisation method. Contour plots are presented to intuitively determine optimal discretisation levels and time steps required for accurate thermal response predictions. This is achieved by comparing numerical solutions of varying discretisation levels with benchmark analytical solutions. Results are reported in terms of governing dimensionless parameters, Biot and Fourier numbers, to ensure generality of findings. Furthermore, spatial and temporal discretisation errors are separated and assessed independently. Finally, models derived using the proposed guidance achieve high levels of prediction accuracy for typically encountered boundary conditions with buildings.