• Korean Chemical Engineering Research
• /
• v.59 no.4
• /
• pp.532-541
• /
• 2021

Recently, research on the application of artificial intelligence in the chemical process has been increasing rapidly. However, overfitting is a significant problem that prevents the model from being generalized well to predict unseen data on test data, as well as observed training data. Cross validation is one of the ways to solve the overfitting problem. In this study, the time-series cross validation method was applied to optimize the number of batch and epoch in the hyperparameters of the prediction model for the 2,3-BDO distillation process, and it compared with K-fold cross validation generally used. As a result, the RMSE of the model with time-series cross validation was lower by 9.06%, and the MAPE was higher by 0.61% than the model with K-fold cross validation. Also, the calculation time was 198.29 sec less than the K-fold cross validation method.

• Journal of Korean Public Health Nursing
• /
• v.28 no.3
• /
• pp.495-508
• /
• 2014

Purpose: This study was conducted in order to evaluate the reliability, validity, sensitivity, and specificity of the Short Form of Bobath Memorial Hospital Fall Risk Assessment Scale (BMFRAS-SF). Methods: A validation study was conducted on 207 elderly patients aged over 65 who were admitted to Bobath Memorial Hospital. Fall risk scores of BMFRAS, composed of eight subscales (age, fall history, physical activity, consciousness level, communication, fall risk factors, underlying disease, and medications) were assessed from the electronic medical record. BMFRAS-SF was derived from eight subscales of the BMFRAS representing the significance between fallers and non-fallers (fall history, physical activity, fall risk factors, underlying disease, and medications). Internal consistency reliability and interrater reliability were assessed by Cronbach's alpha and kappa coefficient. Validity was assessed by Spearman correlation analysis, factor analysis. Sensitivity, specificity, positive predictive and negative predictive values, and a receiver-operating characteristic curve (ROC) were generated. Results: Fallers had significantly higher risk scores than non-fallers in fall history, physical activity, fall risk factors, underlying disease, and medication scales. The BMFRAS-SF demonstrated acceptable Cronbach's alpha (.706) and kappa coefficients of .95. The BMFRAS-SF subscales showed good convergent validity and construct validity. The BMFRAS-SF presented good sensitivity(86.7%), specificity(67.9%), positive predictive value(42.9%) and good negative predictive value(94.8%) at a cut-off score of 5. Areas under the ROC curves were .860 for the BMFRAS and .861 for the BMFRAS-SF. Conclusion: The BMFRAS-SF was proved to be reliable and valid. It could be used for time-saving assessment and evaluation of the high risks for falls in clinical practice settings.

• Journal of Food Hygiene and Safety
• /
• v.34 no.4
• /
• pp.374-379
• /
• 2019

This study developed predictive growth models of Salmonella enterica Serovar Typhimurium on lettuce washed with chlorine (100~300 ppm) and ultrasound (US, 37 kHz, 380 W) treatment and stored at different temperatures ($10{\sim}25^{\circ}C$) using a polynomial equation. The primary model of specific growth rate (SGR) and lag time (LT) showed a good fit ($R^2{\geq}0.92$) with a Gompertz equation. A secondary model was obtained using a quadratic polynomial equation. The appropriateness of the secondary SGR and LT model was verified by coefficient of determination ($R^2=0.98{\sim}0.99$ for internal validation, 0.97~0.98 for external validation), mean square error (MSE=-0.0071~0.0057 for internal validation, -0.0118~0.0176 for external validation), bias factor ($B_f=0.9918{\sim}1.0066$ for internal validation, 0.9865~1.0205 for external validation), and accuracy factor ($A_f=0.9935{\sim}1.0082$ for internal validation, 0.9799~1.0137 for external validation). The newly developed models for S. Typhimurium could be incorporated into a tertiary modeling program to predict the growth of S. Typhimurium as a function of combined chlorine and US during the storage. These new models may also be useful to predict potential S. Typhimurium growth on lettuce, which is important for food safety purposes during the overall supply chain of lettuce from farm to table. Finally, the models may offer reliable and useful information of growth kinetics for the quantification microbial risk assessment of S. Typhimurium on washed lettuce.

• Structural Engineering and Mechanics
• /
• v.47 no.5
• /
• pp.679-700
• /
• 2013

Two new predictive design methods are presented in this study. The first is a hybrid method, called neuro-fuzzy, based on neural networks with fuzzy learning. A total of 280 experimental datasets obtained from the literature concerning concentric punching shear tests of reinforced concrete slab-column connections without shear reinforcement were used to test the model (194 for experimentation and 86 for validation) and were endorsed by statistical validation criteria. The punching shear strength predicted by the neuro-fuzzy model was compared with those predicted by current models of punching shear, widely used in the design practice, such as ACI 318-08, SIA262 and CBA93. The neuro-fuzzy model showed high predictive accuracy of resistance to punching according to all of the relevant codes. A second, more user-friendly design method is presented based on a predictive linear regression model that supports all the geometric and material parameters involved in predicting punching shear. Despite its simplicity, this formulation showed accuracy equivalent to that of the neuro-fuzzy model.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.11 no.4
• /
• pp.293-301
• /
• 2001

This paper reviews three methods to compute a pointwise confidence interval of neuro-fuzzy models and compares their estimation perfonnanee through simulations. The eOITl.putation methods under consideration include stacked generalization using cross-validation, predictive error bar in regressive models, and local reliability measure for the networks employing a local representation scheme. These methods implemented on the neuro-fuzzy models are applied to the problems of simple function approximation and chaotic time series prediction. The results of reliability estimation are compared both quantitatively and qualitatively.

• Journal of Microbiology
• /
• v.43 no.5
• /
• pp.411-416
• /
• 2005

The high-throughput sequencing of microbial genomes has resulted in the relatively rapid accumulation of an enormous amount of genomic sequence data. In this context, the problem posed by the detection of promoters in genomic DNA sequences via computational methods has attracted considerable research attention in recent years. This paper addresses the development of a predictive model, known as the dependence decomposition weight matrix model (DDWMM), which was designed to detect the core promoter region, including the -10 region and the transcription start sites (TSSs), in prokaryotic genomic DNA sequences. This is an issue of some importance with regard to genome annotation efforts. Our predictive model captures the most significant dependencies between positions (allowing for non­adjacent as well as adjacent dependencies) via the maximal dependence decomposition (MDD) procedure, which iteratively decomposes data sets into subsets, based on the significant dependence between positions in the promoter region to be modeled. Such dependencies may be intimately related to biological and structural concerns, since promoter elements are present in a variety of combinations, which are separated by various distances. In this respect, the DDWMM may prove to be appropriate with regard to the detection of core promoter regions and TSSs in long microbial genomic contigs. In order to demonstrate the effectiveness of our predictive model, we applied 10-fold cross-validation experiments on the 607 experimentally-verified promoter sequences, which evidenced good performance in terms of sensitivity.

• International Journal of Highway Engineering
• /
• v.14 no.5
• /
• pp.189-199
• /
• 2012

PURPOSES: The purpose of this study is to propose a new methodology for developing statistical collision models and to show the validation results of the methodology. METHODS: A new modeling method of introducing variables into the model one by one in a multiplicative form is suggested. A method for choosing explanatory variables to be introduced into the model is explained. A method for determining functional forms for each explanatory variable is introduced as well as a parameter estimating procedure. A model selection method is also dealt with. Finally, the validation results is provided to demonstrate the efficacy of the final models developed using the method suggested in this study. RESULTS: According to the results of the validation for the total and injury collisions, the predictive powers of the models developed using the method suggested in this study were better than those of generalized linear models for the same data. CONCLUSIONS: Using the methodology suggested in this study, we could develop better statistical collision models having better predictive powers. This was because the methodology enabled us to find the relationships between dependant variable and each explanatory variable individually and to find the functional forms for the relationships which can be more likely non-linear.

• Journal of Electrical Engineering and Technology
• /
• v.10 no.1
• /
• pp.227-242
• /
• 2015

This paper proposes a weighting factor optimization method in predictive control algorithm for torque ripple reduction in an induction motor fed by an indirect matrix converter (IMC). In this paper, the torque ripple behavior is analyzed to validate the proposed weighting factor optimization method in the predictive control platform and shows the effectiveness of the system. Therefore, an optimization method is adopted here to calculate the optimum weighting factor corresponds to minimum torque ripple and is compared with the results of conventional weighting factor based predictive control algorithm. The predictive control algorithm selects the optimum switching state that minimizes a cost function based on optimized weighting factor to actuate the indirect matrix converter. The conventional and introduced weighting factor optimization method in predictive control algorithm are validated through simulations and experimental validation in DS1104 R&D controller platform and show the potential control, tracking of variables with their respective references and consequently reduces the torque ripple.

• Journal of Institute of Control, Robotics and Systems
• /
• v.14 no.12
• /
• pp.1212-1217
• /
• 2008

This paper studies the feasibility of using the nonlinear model predictive control as a formation flight control algorithm for unmanned aerial vehicles. The optimal control inputs for formation flight are calculated through the cost function which incorporates the relative positions of the individual vehicles to maintain a desired formation and also the inequality constraints on inputs and states using the Karush-Kuhn-Tucker conditions. In the nonlinear model predictive control setting, the optimal control inputs are implemented in a receding horizon manner, which is suitable for dealing with dynamic constraints. Numerical simulations are executed for the validation of the proposed scheme.

• Fire Science and Engineering
• /
• v.24 no.5
• /
• pp.60-67
• /
• 2010

Fire model shall be verified and validated to reliably predict the consequences of fires within its limitations. Generally the verification and validation procedures are conducted by comparison with experimental test data. This study aims to evaluate predictive capabilities of FDS in the pool fire with two rooms and the sensitivity between input parameters such as heat release rate and ventilation rate and the output values like temperature, concentration, and heat flux. The predictive capabilities of FDS will be evaluated by comparing FDS simulation results with PRISME experimental data which result from the international fire test project. The sensitivity analysis will be conducted to decide which one of input parameters affects outcomes by comparison of FDS results with ${\pm}$ 10% changes of input parameter. From this study, the FDS predictive capabilities are within 20% error range. Heat release rate as input parameter affects most of outcomes and flow rate only has relation with concentration of oxygen and combustion products.