• Smart Media Journal
• /
• v.13 no.2
• /
• pp.9-15
• /
• 2024

In this paper, a Diffusion Multi-step Classifier (DMC) is proposed to address the imbalance issue in credit prediction. DMC utilizes a Diffusion Model to generate continuous numerical data from credit prediction data and creates categorical data through a Multi-step Classifier. Compared to other algorithms generating synthetic data, DMC produces data with a distribution more similar to real data. Using DMC, data that closely resemble actual data can be generated, outperforming other algorithms for data generation. When experiments were conducted using the generated data, the probability of predicting delinquencies increased by over 20%, and overall predictive accuracy improved by approximately 4%. These research findings are anticipated to significantly contribute to reducing delinquency rates and increasing profits when applied in actual financial institutions.

• Communications for Statistical Applications and Methods
• /
• v.31 no.4
• /
• pp.441-457
• /
• 2024

Semicontinuous data are characterized by a mixture of a point probability mass at zero and a continuous distribution of positive values. This type of data is often modeled using a two-part model where the first part models the probability of dichotomous outcomes -zero or positive- and the second part models the distribution of positive values. Despite the two-part model's popularity, variable selection in this model has not been fully addressed, especially, in high dimensional data. The objective of this study is to investigate variable selection and prediction performance of penalized regression methods in two-part models. The performance of the selected techniques in the two-part model is evaluated via simulation studies. Our findings show that LASSO and ENET tend to select more predictors in the model than SCAD and MCP. Consequently, MCP and SCAD outperform LASSO and ENET for β-specificity, and LASSO and ENET perform better than MCP and SCAD with respect to the mean squared error. We find similar results when applying the penalized regression methods to the prediction of crime incidents using community-based data.

• Smart Structures and Systems
• /
• v.30 no.6
• /
• pp.601-612
• /
• 2022

The prediction of structural mechanical behaviors is vital important to early perceive the abnormal conditions and avoid the occurrence of disasters. Especially for underground engineering, complex geological conditions make the structure more prone to disasters. Aiming at solving the problems existing in previous studies, such as incomplete consideration factors and can only predict the continuous performance, the deep attention fused temporal convolution network (DATCN) is proposed in this paper to predict the spatial mechanical behaviors of structure, which integrates both the temporal effect and spatial effect and realize the cross-time prediction. The temporal convolution network (TCN) and self-attention mechanism are employed to learn the temporal correlation of each monitoring point and the spatial correlation among different points, respectively. Then, the predicted result obtained from DATCN is compared with that obtained from some classical baselines, including SVR, LR, MLP, and RNNs. Also, the parameters involved in DATCN are discussed to optimize the prediction ability. The prediction result demonstrates that the proposed DATCN model outperforms the state-of-the-art baselines. The prediction accuracy of DATCN model after 24 hours reaches 90 percent. Also, the performance in last 14 hours plays a domain role to predict the short-term behaviors of the structure. As a study case, the proposed model is applied in an underwater shield tunnel to predict the stress variation of concrete segments in space.

• Journal of the Korean Data Analysis Society
• /
• v.20 no.6
• /
• pp.2829-2840
• /
• 2018

The random variable with an arbitrary value or more is called semi-continuous variable or zero-inflated one in case that its boundary value is more frequently observed than expected. This means the boundary value is likely to be practically observed more than it should be theoretically under certain probability distribution. When the distribution considered is continuous, the variable is defined as semi-continuous and when one of discrete distribution is assumed for the variable, we regard it as zero-inflated. In this study, we introduce the two-part model, which consists of one part for modelling the binary response and the other part for modelling the variable greater than the boundary value. Especially, the zero-inflated regression models are explained by using Poisson distribution and negative binomial distribution. In real data analysis, we employ the zero-inflated regression models to estimate the number of days under extreme heat-wave circumstances during the last 10 years in South Korea. Based on the estimation results, we create prediction maps for the estimated number of days under heat-wave advisory and heat-wave warning by using the universal kriging, which is one of the spatial prediction methods.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.37 no.6
• /
• pp.981-987
• /
• 2017

Jeju Island is a volcanic island which has a large permeability. Groundwater is a major water resources and its proper management is essential. Especially, there is a multilevel restriction due to the groundwater level decline during a drought period to protect sea water intrusion. Preliminary countermeasure using long-term groundwater level prediction is necessary to use agricultural groundwater properly. For this purpose, the monthly groundwater level prediction technique by Artificial Neural Network model was developed and applied to the representative monitoring wells. The monthly prediction model showed excellent results for training and test periods. The continuous groundwater level prediction model also developed, which used the monthly forecasted values adaptively as input data. The characteristics of groundwater declines were analyzed under extreme cases without precipitation for several months.

• KIPS Transactions on Software and Data Engineering
• /
• v.2 no.4
• /
• pp.281-290
• /
• 2013

The IT asset is a core part that supports the management objective of an organization, and the fast settlement of the IT asset fault is very important. In this study, a fault recovery prediction technique is proposed, which uses the existing fault data to address the IT asset fault. The proposed fault recovery prediction technique is as follows. First, the existing fault recovery data were pre-processed and classified by fault recovery type; second, a rule was established for the keyword mapping of the classified fault recovery types and reported data; and third, a machine learning process that allows the prediction of the fault recovery method based on the established rule was presented. To verify the effectiveness of the proposed machine learning process, company A's 33,000 computer fault data for the duration of six months were tested. The hit rate for fault recovery prediction was approximately 72%, and it increased to 81% via continuous machine learning.

• Journal of Soil and Groundwater Environment
• /
• v.17 no.5
• /
• pp.56-67
• /
• 2012

In the present study, an enhanced subsurface prediction algorithm based on a non-parametric geostatistical model and a history matching technique through Gibbs sampler is developed and the iterative prediction improvement procedure is proposed. The developed model is applied to a simple two-dimensional synthetic case where domain is composed of three different hydrogeologic media with $500m{\times}40m$ scale. In the application, it is assumed that there are 4 independent pumping tests performed at different vertical interval and the history curves are acquired through numerical modeling. With two hypothetical borehole information and pumping test data, the proposed prediction model is applied iteratively and continuous improvements of the predictions with reduced uncertainties of the media distribution are observed. From the results and the qualitative/quantitative analysis, it is concluded that the proposed model is good for the subsurface prediction improvements where the history data is available as a supportive information. Once the proposed model be a matured technique, it is believed that the model can be applied to many groundwater, geothermal, gas and oil problems with conventional fluid flow simulators. However, the overall development is still in its preliminary step and further considerations needs to be incorporated to be a viable and practical prediction technique including multi-dimensional verifications, global optimization, etc. which have not been resolved in the present study.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2004.10a
• /
• pp.25-28
• /
• 2004

A fracture was generated by change of clearance and deterioration of material properties on the sheet metal through temperature. This paper describes the results of a prediction about the temperature of the sheet metal during continuous stamping process, because the temperature increase of the sheet metal has a detrimental effect on formability. To analyze the temperature increase of the sheet metal during continuous stamping process, tensile and friction tests were performed from room temperature to 300$^{\circ}C$ at warm condition in this study. As temperature increase, tensile strength, elongation, strain hardening exponent and anisotropy coefficient for each specimens were decreased. On the other hand, friction coefficients were increased. From the FE-simulation results, temperature upward tendency was identified on dies and sheet metal. These observations are rationalized on the basis of the material properties, friction coefficient vs. temperature relationship for the sheet.

• Clinical and Experimental Pediatrics
• /
• v.61 no.10
• /
• pp.307-314
• /
• 2018

The incidence of type 1 diabetes mellitus (T1DM) in children and adolescents is increasing worldwide. Combined effects of genetic and environmental factors cause T1DM, which make it difficult to predict whether an individual will inherit the disease. Due to the level of self-care necessary in T1DM maintenance, it is crucial for pediatric settings to support achieving optimal glucose control, especially when adolescents are beginning to take more responsibility for their own health. Innovative insulin delivery systems, such as continuous subcutaneous insulin infusion (CSII), and noninvasive glucose monitoring systems, such as continuous glucose monitoring (CGM), allow patients with T1DM to achieve a normal and flexible lifestyle. However, there are still challenges in achieving optimal glucose control despite advanced technology in T1DM administration. In this article, disease prediction and current management of T1DM are reviewed with special emphasis on biomarkers of pancreatic ${\beta}-cell$ stress, CSII, glucose monitoring, and several other adjunctive therapies.

• Clinical and Experimental Pediatrics
• /
• v.52 no.9
• /
• pp.964-970
• /
• 2009

Neonatal seizures are generally not only brief and subtle but also not easily recognized and are usually untreated. In sick neonates, seizures are frequently not manifested clinically but are detected only by electroencephalography (subclinical EEG seizures). This phenomenon of electroclinical dissociation is fairly common in neonates. On the other hand, neonates frequently show clinical behaviors such as stiffening, apnea, or autonomic manifestations that mimic seizures, which is usually associated with underlying encephalopathy and non-epileptic seizures. Therefore, it might be difficult to confirm the diagnosis of neonatal seizures. Early recognition of neonatal seizures is important to minimize poor neurodevelopmental outcomes, including cognitive, behavioral, and learning disabilities, as well as the development of postnatal epilepsy. EEG is a reliable tool in the determination of neonatal seizures. Continuous EEG monitoring is essential for the identification of seizures, evaluation of treatment efficacy, and prediction of the neurodevelopmental outcome. However, there is not yet a wide consensus on the optimal "standard" lead montage for the continuous EEG monitoring.