• Journal of Intelligence and Information Systems
• /
• v.25 no.1
• /
• pp.127-137
• /
• 2019

Recidivism prediction has been a subject of constant research by experts since the early 1970s. But it has become more important as committed crimes by recidivist steadily increase. Especially, in the 1990s, after the US and Canada adopted the 'Recidivism Risk Assessment Report' as a decisive criterion during trial and parole screening, research on recidivism prediction became more active. And in the same period, empirical studies on 'Recidivism Factors' were started even at Korea. Even though most recidivism prediction studies have so far focused on factors of recidivism or the accuracy of recidivism prediction, it is important to minimize the prediction misclassification cost, because recidivism prediction has an asymmetric error cost structure. In general, the cost of misrecognizing people who do not cause recidivism to cause recidivism is lower than the cost of incorrectly classifying people who would cause recidivism. Because the former increases only the additional monitoring costs, while the latter increases the amount of social, and economic costs. Therefore, in this paper, we propose an XGBoost(eXtream Gradient Boosting; XGB) based recidivism prediction model considering asymmetric error cost. In the first step of the model, XGB, being recognized as high performance ensemble method in the field of data mining, was applied. And the results of XGB were compared with various prediction models such as LOGIT(logistic regression analysis), DT(decision trees), ANN(artificial neural networks), and SVM(support vector machines). In the next step, the threshold is optimized to minimize the total misclassification cost, which is the weighted average of FNE(False Negative Error) and FPE(False Positive Error). To verify the usefulness of the model, the model was applied to a real recidivism prediction dataset. As a result, it was confirmed that the XGB model not only showed better prediction accuracy than other prediction models but also reduced the cost of misclassification most effectively.

• Korean Journal of Food Science and Technology
• /
• v.47 no.1
• /
• pp.95-102
• /
• 2015

The Maillard reaction is a non-enzymatic reaction between amino and carbonyl groups. During milk processing, lactose reacts with milk protein through this reaction. Infant formulas (IFs) are milk-based products processed with heat-treatments, including spray-drying and sterilization. Because IFs contain higher Maillard reaction products (MRPs) than breast milk, formula-fed infants are subject to higher MRP exposure than breast milk-fed ones. In this study, we investigated the optimization of conditions for minimal MRP formation with the addition of $\small{L}$-carnitine ($\small{L}$-car), pyridoxine hydrochloride (PH), and $\small{DL}$-${\alpha}$-tocopheryl acetate (${\alpha}$-T) in an IF model system. MRP formation was monitored by response surface methodology using fluorescence intensity (FI) and 5-hydroxymethylfurfural (HMF) content. The optimal condition for minimizing the formation of MRPs was with $2.3{\mu}M$ $\small{L}$-car, $15.8{\mu}M$ PH, and $20.6{\mu}M$ ${\alpha}$-T. Under this condition, the predicted values were 77.4% FI and 248.7 ppb HMF.

• Journal of Intelligence and Information Systems
• /
• v.17 no.4
• /
• pp.241-254
• /
• 2011

Financial time-series forecasting is one of the most important issues because it is essential for the risk management of financial institutions. Therefore, researchers have tried to forecast financial time-series using various data mining techniques such as regression, artificial neural networks, decision trees, k-nearest neighbor etc. Recently, support vector machines (SVMs) are popularly applied to this research area because they have advantages that they don't require huge training data and have low possibility of overfitting. However, a user must determine several design factors by heuristics in order to use SVM. For example, the selection of appropriate kernel function and its parameters and proper feature subset selection are major design factors of SVM. Other than these factors, the proper selection of instance subset may also improve the forecasting performance of SVM by eliminating irrelevant and distorting training instances. Nonetheless, there have been few studies that have applied instance selection to SVM, especially in the domain of stock market prediction. Instance selection tries to choose proper instance subsets from original training data. It may be considered as a method of knowledge refinement and it maintains the instance-base. This study proposes the novel instance selection algorithm for SVMs. The proposed technique in this study uses genetic algorithm (GA) to optimize instance selection process with parameter optimization simultaneously. We call the model as ISVM (SVM with Instance selection) in this study. Experiments on stock market data are implemented using ISVM. In this study, the GA searches for optimal or near-optimal values of kernel parameters and relevant instances for SVMs. This study needs two sets of parameters in chromosomes in GA setting : The codes for kernel parameters and for instance selection. For the controlling parameters of the GA search, the population size is set at 50 organisms and the value of the crossover rate is set at 0.7 while the mutation rate is 0.1. As the stopping condition, 50 generations are permitted. The application data used in this study consists of technical indicators and the direction of change in the daily Korea stock price index (KOSPI). The total number of samples is 2218 trading days. We separate the whole data into three subsets as training, test, hold-out data set. The number of data in each subset is 1056, 581, 581 respectively. This study compares ISVM to several comparative models including logistic regression (logit), backpropagation neural networks (ANN), nearest neighbor (1-NN), conventional SVM (SVM) and SVM with the optimized parameters (PSVM). In especial, PSVM uses optimized kernel parameters by the genetic algorithm. The experimental results show that ISVM outperforms 1-NN by 15.32%, ANN by 6.89%, Logit and SVM by 5.34%, and PSVM by 4.82% for the holdout data. For ISVM, only 556 data from 1056 original training data are used to produce the result. In addition, the two-sample test for proportions is used to examine whether ISVM significantly outperforms other comparative models. The results indicate that ISVM outperforms ANN and 1-NN at the 1% statistical significance level. In addition, ISVM performs better than Logit, SVM and PSVM at the 5% statistical significance level.

• KSBB Journal
• /
• v.23 no.6
• /
• pp.463-469
• /
• 2008

During the last decade, monacolin-K biosynthesized by fermentation of red yeast rice (Monascus strains) was proved to have an efficient cholesterol lowering capability, leading to rapid increase in the market demand for the functional red yeast rice. In this study, the production medium composition and components were optimized on a shake flask scale for monacolin-K production by Monascus pilosus (KCCM 60160). The effect of three different soybean flours on the monacolin-K production were studied in order to replace the nitrogen sources of basic production medium (yeast extract, malt extract and beef extract). Among the several experiments, the production medium with dietary soybean flour to replace a half of yeast extract was very good for monacolin-K production. Plackett-Burman experimental design was used to determine the key factors which are critical to produce the biological products in the fermentation. According to the result of Plackett-Burman experimental design, a second order response surface design was applied using yeast extract, beef extract and $(NH_4)_2SO_4$ as factors. Applying this model, the optimum concentration of the three variables was obtained. The maximum monacolin-K production (369.6 mg/L) predicted by model agrees well with the experimental value (418 mg/L) obtained from the experimental verification at the optimal medium. The yield of monacolin-K was increased by 67% as compared to that obtained with basic production medium in shake flasks.

• Journal of Intelligence and Information Systems
• /
• v.22 no.2
• /
• pp.127-142
• /
• 2016

Deep learning model is a kind of neural networks that allows multiple hidden layers. There are various deep learning architectures such as convolutional neural networks, deep belief networks and recurrent neural networks. Those have been applied to fields like computer vision, automatic speech recognition, natural language processing, audio recognition and bioinformatics where they have been shown to produce state-of-the-art results on various tasks. Among those architectures, convolutional neural networks and recurrent neural networks are classified as the supervised learning model. And in recent years, those supervised learning models have gained more popularity than unsupervised learning models such as deep belief networks, because supervised learning models have shown fashionable applications in such fields mentioned above. Deep learning models can be trained with backpropagation algorithm. Backpropagation is an abbreviation for "backward propagation of errors" and a common method of training artificial neural networks used in conjunction with an optimization method such as gradient descent. The method calculates the gradient of an error function with respect to all the weights in the network. The gradient is fed to the optimization method which in turn uses it to update the weights, in an attempt to minimize the error function. Convolutional neural networks use a special architecture which is particularly well-adapted to classify images. Using this architecture makes convolutional networks fast to train. This, in turn, helps us train deep, muti-layer networks, which are very good at classifying images. These days, deep convolutional networks are used in most neural networks for image recognition. Convolutional neural networks use three basic ideas: local receptive fields, shared weights, and pooling. By local receptive fields, we mean that each neuron in the first(or any) hidden layer will be connected to a small region of the input(or previous layer's) neurons. Shared weights mean that we're going to use the same weights and bias for each of the local receptive field. This means that all the neurons in the hidden layer detect exactly the same feature, just at different locations in the input image. In addition to the convolutional layers just described, convolutional neural networks also contain pooling layers. Pooling layers are usually used immediately after convolutional layers. What the pooling layers do is to simplify the information in the output from the convolutional layer. Recent convolutional network architectures have 10 to 20 hidden layers and billions of connections between units. Training deep learning networks has taken weeks several years ago, but thanks to progress in GPU and algorithm enhancement, training time has reduced to several hours. Neural networks with time-varying behavior are known as recurrent neural networks or RNNs. A recurrent neural network is a class of artificial neural network where connections between units form a directed cycle. This creates an internal state of the network which allows it to exhibit dynamic temporal behavior. Unlike feedforward neural networks, RNNs can use their internal memory to process arbitrary sequences of inputs. Early RNN models turned out to be very difficult to train, harder even than deep feedforward networks. The reason is the unstable gradient problem such as vanishing gradient and exploding gradient. The gradient can get smaller and smaller as it is propagated back through layers. This makes learning in early layers extremely slow. The problem actually gets worse in RNNs, since gradients aren't just propagated backward through layers, they're propagated backward through time. If the network runs for a long time, that can make the gradient extremely unstable and hard to learn from. It has been possible to incorporate an idea known as long short-term memory units (LSTMs) into RNNs. LSTMs make it much easier to get good results when training RNNs, and many recent papers make use of LSTMs or related ideas.

• Journal of Life Science
• /
• v.32 no.11
• /
• pp.872-881
• /
• 2022

In this study, we optimized the composition of the indole-3-acetic acid (IAA) production medium using response surface methodology on Pantoea agglomerans SRCM 119864 isolated from soil. IAA-producing P. aglomerans SRCM 119864 was identified by 16S rRNA gene sequencing. There are 11 intermediate components known to affect IAA production, hence the effect of each component on IAA production was investigated using a Plackett-Burman design (PBD). Based on the PBD, sucrose, tryptone, and sodium chloride were selected as the main factors that enhanced the IAA production at optimal L-tryptophan concentration. The predicted maximum IAA production (64.34 mg/l) was obtained for a concentration of sucrose of 13.38 g/l, of tryptone of 18.34 g/l, of sodium chloride of 9.71 g/l, and of L-tryptophan of 6.25 g/l using a the hybrid design experimental model. In the experiment, the nutrient broth medium supplemented with 0.1% L-tryptophan as the basal medium produced 45.24 mg/l of IAA, whereas the optimized medium produced 65.40 mg/l of IAA, resulting in a 44.56% increase in efficiency. It was confirmed that the IAA production of the designed optimal composition medium was very similar to the predicted IAA production. The statistical significance and suitability of the experimental model were verified through analysis of variance (ANOVA). Therefore, in this study, we determined the optimal growth medium concentration for the maximum production of IAA, which can contribute to sustainable agriculture and increase crop yield.

• Clean Technology
• /
• v.28 no.3
• /
• pp.218-226
• /
• 2022

As a result of the recent social transformation towards a hydrogen economy and carbon-neutrality, the demands for hydrogen energy have been increasing rapidly worldwide. As such, eco-friendly hydrogen production technologies that do not produce carbon dioxide (CO₂) emissions are being focused on. Among them, ammonia (NH₃) is an economical hydrogen carrier that can easily produce hydrogen (H₂). In this study, Ru/Al₂O₃ catalyst coated onmetallic monolith for hydrogen production from ammonia was prepared by a dip-coating method using a catalyst slurry mixture composed of Ru/Al₂O₃ catalyst, inorganic binder (alumina sol) and organic binder (methyl cellulose). At the optimized 1:1:0.1 weight ratio of catalyst/inorganic binder/organic binder, the amount of catalyst coated on the metallic monolith after one cycle coating was about 61.6 g L^-1. The uniform thickness (about 42 ㎛) and crystal structure of the catalyst coated on the metallic monolith surface were confirmed through scanning electron microscopy (SEM) and X-ray diffraction (XRD) analysis. Also, a numerical optimization regression equation for NH3 conversion according to the independent variables of reaction temperature (400-600 ℃) and gas hourly space velocity (1,000-5,000 h^-1) was calculated by response surface methodology (RSM). This model indicated a determination coefficient (R²) of 0.991 and had statistically significant predictors. This regression model could contribute to the commercial process design of hydrogen production by ammonia decomposition.

• Journal of the Korean GEO-environmental Society
• /
• v.18 no.2
• /
• pp.47-51
• /
• 2017

In this study, the optimization algorithm by inverse analysis that is the core of the adaptive management technique was adopted to update the soil engineering properties based on the ground response during the construction. Adaptive management technique is the framework wherein construction and design procedures are adjusted based on observations and measurements made as construction proceeds. To evaluate the performance of the adaptive management technique, the numerical simulation for the triaxial tests and the synthetic deep excavation were conducted with the Hardening Soil model. To effectively conduct the analysis, the effective parameters among the parameters employed in the model were selected based on the composite scaled sensitivity analysis. The results from the undrained triaxial tests performed with soft Chicago clays were used for the parameter calibration. The simulation for the synthetic deep excavation were conducted assuming that the soil engineering parameters obtained from the triaxial simulation represent the actual field condition. These values were used as the reference values. The observation for the synthetic deep excavation simulations was the horizontal displacement of the support wall that has the highest composite scaled sensitivity among the other possible observations. It was found that the horizontal displacement of the support wall with the various initial soil properties were converged to the reference displacement by using the adaptive management technique.

• Journal of the Korean Society of Hazard Mitigation
• /
• v.8 no.5
• /
• pp.53-58
• /
• 2008

This study presented a design model optimizing a distance of inlet with drainage pipe laid under the gutter in road. When the distance of inlet changed, a basin for the gutter divided by the distance of inlet and the inflow coming into the gutter would be changed. In this case, the change of inlet distance causes the change of a diameter of drainage pipe and slope because of the change of capacity. Therefore, the optimization is needed to design the combination of them for the distance of inlet. Genetic Algorithm is used to determine the optimal combination of them. The conditions of road and the precipitation were assumed like a real and the range of inlet distance adopted $10{\sim}30\;m$ which has been introduced in domestic. This model presented the optimal distance of inlet and the combination of pipe and slope through the minimum cost. The result of the study is that the optimal distance of inlet is different from each slope of road and it can reduce about 20% of total cost for the distance of inlet.

• Investigative Magnetic Resonance Imaging
• /
• v.21 no.2
• /
• pp.65-70
• /
• 2017

Purpose: To optimize the saturation time and maximizing the pH-weighted difference between the normal and ischemic brain regions, on 3-tesla amide proton transfer (APT) imaging using an in vivo rat model. Materials and Methods: Three male Wistar rats underwent middle cerebral artery occlusion, and were examined in a 3-tesla magnetic resonance imaging (MRI) scanner. APT imaging acquisition was performed with 3-dimensional turbo spin-echo imaging, using a 32-channel head coil and 2-channel parallel radiofrequency transmission. An off-resonance radiofrequency pulse was applied with a Sinc-Gauss pulse at a $B_{1,rms}$ amplitude of $1.2{\mu}T$ using a 2-channel parallel transmission. Saturation times of 3, 4, or 5 s were tested. The APT effect was quantified using the magnetization-transfer-ratio asymmetry at 3.5 ppm with respect to the water resonance (APT-weighted signal), and compared with the normal and ischemic regions. The result was then applied to an acute stroke patient to evaluate feasibility. Results: Visual detection of ischemic regions was achieved with the 3-, 4-, and 5-s protocols. Among the different saturation times at $1.2{\mu}T$ power, 4 s showed the maximum difference between the ischemic and normal regions (-0.95%, P = 0.029). The APTw signal difference for 3 and 5 s was -0.9% and -0.7%, respectively. The 4-s saturation time protocol also successfully depicted the pH-weighted differences in an acute stroke patient. Conclusion: For 3-tesla turbo spin-echo APT imaging, the maximal pH-weighted difference achieved when using the $1.2{\mu}T$ power, was with the 4 s saturation time. This protocol will be helpful to depict pH-weighted difference in stroke patients in clinical settings.