• Journal of Industrial Technology
• /
• v.39 no.1
• /
• pp.27-31
• /
• 2019

Semiconductor logistics systems are facing difficulties in increasing production as production processes become more complicated due to the upgrading of fine processes. Therefore, the purpose of the research is to design predictive models that can predict traffic during the pre-planning stage, identify the risk zones that occur during the production process, and prevent them in advance. As a solution, we build FABs using automode simulation to collect data. Then, the traffic prediction model of the areas of interest is constructed using deep learning techniques (keras - multistory conceptron structure). The design of the predictive model gave an estimate of the traffic in the area of interest with an accuracy of about 87%. The expected effect can be used as an indicator for making decisions by proactively identifying congestion risk areas during the Fab Design or Factory Expansion Planning stage, as the maximum traffic per section is predicted.

• Korean Journal of Artificial Intelligence
• /
• v.11 no.3
• /
• pp.29-34
• /
• 2023

Recently, the rapid development of artificial intelligence technology, many studies are being conducted to predict the risk of heart disease in order to lower the mortality rate of cardiovascular diseases worldwide. This study presents exercise or dietary improvement contents in the form of a software app or web to patients with cardiovascular disease, and cardiovascular disease through digital devices such as mobile phones and PCs. LR, LDA, SVM, XGBoost for the purpose of developing "Life style Improvement Contents (Digital Therapy)" for cardiovascular disease care to help with management or treatment We compared and analyzed cardiovascular disease prediction models using machine learning algorithms. Research Results XGBoost. The algorithm model showed the best predictive model performance with overall accuracy of 80% before and after. Overall, accuracy was 80.0%, F1 Score was 0.77~0.79, and ROC-AUC was 80%~84%, resulting in predictive model performance. Therefore, it was found that the algorithm used in this study can be used as a reference model necessary to verify the validity and accuracy of cardiovascular disease prediction. A cardiovascular disease prediction analysis algorithm that can enter accurate biometric data collected in future clinical trials, add lifestyle management (exercise, eating habits, etc.) elements, and verify the effect and efficacy on cardiovascular-related bio-signals and disease risk. development, ultimately suggesting that it is possible to develop lifestyle improvement contents (Digital Therapy).

• Structural Engineering and Mechanics
• /
• v.91 no.3
• /
• pp.279-289
• /
• 2024

The purpose of this study is to represent a useful alternative for the preliminary seismic vulnerability assessment of existing reinforced concrete buildings by introducing a statistical approach employing the binary logistic regression technique. Two different predictive statistical models, namely full and reduced models, were generated utilizing building characteristics obtained from the damage database compiled after 1999 Düzce earthquake. Among the inspected building parameters, number of stories, overhang ratio, priority index, soft story index, normalized redundancy ratio and normalized lateral stiffness index were specifically selected as the predictor variables for vulnerability classification. As a result, normalized redundancy ratio and soft story index were identified as the most significant predictors affecting seismic vulnerability in terms of life safety performance level. In conclusion, it is revealed that both models are capable of classifying the set of buildings being severely damaged or collapsed with a balanced accuracy of 73%, hence, both are able to filter out high-priority buildings for life safety performance assessment. Thus, in this study, having the same high accuracy as the full model, the reduced model using fewer predictors is proposed as a simple and viable classifier for determining life safety levels of reinforced concrete buildings in the preliminary seismic risk assessment.

• Asian Pacific Journal of Cancer Prevention
• /
• v.15 no.2
• /
• pp.867-870
• /
• 2014

Purpose: This study used receiver operating characteristic curve to analyze Surveillance, Epidemiology and End Results (SEER) ependymoma data to identify predictive models and potential disparity in outcome. Materials and Methods: This study analyzed socio-economic, staging and treatment factors available in the SEER database for ependymoma. For the risk modeling, each factor was fitted by a Generalized Linear Model to predict the outcome ('brain and other nervous systems' specific death in yes/no). The area under the receiver operating characteristic curve (ROC) was computed. Similar strata were combined to construct the most parsimonious models. A random sampling algorithm was used to estimate the modeling errors. Risk of ependymoma death was computed for the predictors for comparison. Results: A total of 3,500 patients diagnosed from 1973 to 2009 were included in this study. The mean follow up time (S.D.) was 79.8 (82.3) months. Some 46% of the patients were female. The mean (S.D.) age was 34.4 (22.8) years. Age was the most predictive factor of outcome. Unknown grade demonstrated a 15% risk of cause specific death compared to 9% for grades I and II, and 36% for grades III and IV. A 5-tiered grade model (with a ROC area 0.48) was optimized to a 3-tiered model (with ROC area of 0.53). This ROC area tied for the second with that for surgery. African-American patients had 21.5% risk of death compared with 16.6% for the others. Some 72.7% of patient who did not get RT had cerebellar or spinal ependymoma. Patients undergoing surgery had 16.3% risk of death, as compared to 23.7% among those who did not have surgery. Conclusion: Grading ependymoma may dramatically improve modeling of data. RT is under used for cerebellum and spinal cord ependymoma and it may be a potential way to improve outcome.

• Journal of The Geomorphological Association of Korea
• /
• v.27 no.2
• /
• pp.65-80
• /
• 2020

This paper deals with risk assessment of life in a landslide-prone area by a GIS-based modeling method. Landslide susceptibility maps can provide a probability of landslide prone areas to mitigate or proper control this problems and to take any development plan and disaster management. A landslide inventory map of the study area was prepared based on past historical information and aerial photography analysis. A total of 550 landslides have been counted at the whole study area. The extracted landslides were randomly selected and divided into two different groups, 50% of the landslides were used for model calibration and the other were used for validation purpose. Eleven causative factors (continuous and thematic) such as slope, aspect, curvature, topographic wetness index, elevation, forest type, forest crown density, geology, land-use, soil drainage, and soil texture were used in hazard analysis. The correlation between landslides and these factors, pixels were divided into several classes and frequency ratio was also extracted. Eventually, a landslide susceptibility map was constructed using a logistic regression model based on entire events. Moreover, the landslide susceptibility map was plotted with a receiver operating characteristic (ROC) curve and calculated the area under the curve (AUC) and tried to extract a success rate curve. Based on the results, logistic regression produced an 85.18% accuracy, so we believed that the model was reliable and acceptable for the landslide susceptibility analysis on the study area. In addition, for risk assessment, vulnerability scale were added for social thematic data layer. The study area predictive landslide affected pixels 2,000 and 5,000 were also calculated for making a probability table. In final calculation, the 2,000 predictive landslide affected pixels were assumed to run. The total population causalities were estimated as 7.75 person that was relatively close to the actual number published in Korean Annual Disaster Report, 2006.

• Journal of Preventive Medicine and Public Health
• /
• v.32 no.4
• /
• pp.538-545
• /
• 1999

Objective : To compare the predictive power of International Classification of Diseases 10th Edition(ICD-10) based International Classification of Diseases based Injury Severity Score(ICISS) with Trauma and Injury Severity Score(TRISS) and International Classification of Diseases 9th Edition Clinical Modification(ICD-9CM) based ICISS in the injury severity measure. Methods : ICD-10 version of Survival Risk Ratios(SRRs) was derived from 47,750 trauma patients from 35 Emergency Centers for 1 year. The predictive power of TRISS, the ICD-9CM based ICISS and ICD-10 based ICISS were compared in a group of 367 severely injured patients admitted to two university hospitals. The predictive power was compared by using the measures of discrimination(disparity, sensitivity, specificity, misclassification rates, and ROC curve analysis) and calibration(Hosmer-Lemeshow goodness-of-fit statistics), all calculated by logistic regression procedure. Results : ICD-10 based ICISS showed a lower performance than TRISS and ICD-9CM based ICISS. When age and Revised Trauma Score(RTS) were incorporated into the survival probability model, however, ICD-10 based ICISS full model showed a similar predictive power compared with TRISS and ICD-9CM based ICISS full model. ICD-10 based ICISS had some disadvantages in predicting outcomes among patients with intracranial injuries. However, such weakness was largely compensated by incorporating age and RTS in the model. Conclusions : The ICISS methodology can be extended to ICD-10 horizon as a standard injury severity measure in the place of TRISS, especially when age and RTS were incorporated in the model. In patients with intracranial injuries, the predictive power of ICD-10 based ICISS was relatively low because of differences in the classifying system between ICD-10 and ICD-9CM.

• Journal of Food Hygiene and Safety
• /
• v.33 no.3
• /
• pp.157-161
• /
• 2018

The objective of this study was to develop software to predict the kinetic behavior and the probability of foodborne bacterial growth on processed meat products. It is designed for rapid application by non-specialists in predictive microbiology. The software, named Foodborne bacteria Animal product Modeling Equipment (FAME), was developed using Javascript and HTML. FAME consists of a kinetic model and a probabilistic model, and it can be used to predict bacterial growth pattern and probability. In addition, validation and editing of model equation are available in FAME. The data used by the software were constructed with 5,400 frankfurter samples for the kinetic model and 345,600 samples for the probabilistic model using a variety of combinations including atmospheric conditions, temperature, NaCl concentrations and $NaNO_2$ concentrations. Using FAME, users can select the concentrations of NaCl and $NaNO_2$ meat products as well as storage conditions (atmosphere and temperature). The software displays bacterial growth patterns and growth probabilities, which facilitate the determination of optimal safety conditions for meat products. FAME is useful in predicting bacterial kinetic behavior and growth probability, especially for quick application, and is designed for use by non-specialists in predictive microbiology.

• Journal of information and communication convergence engineering
• /
• v.21 no.1
• /
• pp.82-89
• /
• 2023

Various machine-learning models may yield high predictive power for massive time series for time series prediction. However, these models are prone to instability in terms of computational cost because of the high dimensionality of the feature space and nonoptimized hyperparameter settings. Considering the potential risk that model training with a high-dimensional feature set can be time-consuming, we evaluate a feature-importance-based feature selection method to derive a tradeoff between predictive power and computational cost for time series prediction. We used two machine learning techniques for performance evaluation to generate prediction models from a retail sales dataset. First, we ranked the features using impurity- and Local Interpretable Model-agnostic Explanations (LIME) -based feature importance measures in the prediction models. Then, the recursive feature elimination method was applied to eliminate unimportant features sequentially. Consequently, we obtained a subset of features that could lead to reduced model training time while preserving acceptable model performance.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.22 no.2
• /
• pp.51-62
• /
• 2022

This study suggests a design of predictive modeling for a hospital fall risk based on inpatients' posture. Inpatient's profile, medical history, and body measurement data along with basic information about a bed they use, were used to predict a fall risk and suggest an algorithm to determine the level of risk. Fall risk prediction is largely divided into two parts: a real-time fall risk evaluation and a qualitative fall risk exposure assessment, which is mostly based on the inpatient's profile. The former is carried out by recognizing an inpatient's posture in bed and extracting rule-based information to measure fall risk while the latter is conducted by medical staff who examines an inpatient's health status related to hospital fall risk and assesses the level of risk exposure. The inpatient fall risk is determined using a sigmoid function with recognized inpatient posture information, body measurement data and qualitative risk assessment results combined. The procedure and prediction model suggested in this study is expected to significantly contribute to tailored services for inpatients and help ensure hospital fall prevention and inpatient safety.

• Asian Pacific Journal of Cancer Prevention
• /
• v.15 no.16
• /
• pp.6781-6785
• /
• 2014

Purpose: This study used receiver operating characteristic curves to analyze Surveillance, Epidemiology and End Results (SEER) medulloblastoma (MB) and primitive neuroectodermal tumor (PNET) outcome data. The aim of this study was to identify and optimize predictive outcome models. Materials and Methods: Patients diagnosed from 1973 to 2009 were selected for analysis of socio-economic, staging and treatment factors available in the SEER database for MB and PNET. For the risk modeling, each factor was fitted by a generalized linear model to predict the outcome (brain cancer specific death, yes/no). The area under the receiver operating characteristic curve (ROC) was computed. Similar strata were combined to construct the most parsimonious models. A Monte Carlo algorithm was used to estimate the modeling errors. Results: There were 3,702 patients included in this study. The mean follow up time (S.D.) was 73.7 (86.2) months. Some 40% of the patients were female and the mean (S.D.) age was 16.5 (16.6) years. There were more adult MB/PNET patients listed from SEER data than pediatric and young adult patients. Only 12% of patients were staged. The SEER staging has the highest ROC (S.D.) area of 0.55 (0.05) among the factors tested. We simplified the 3-layered risk levels (local, regional, distant) to a simpler non-metastatic (I and II) versus metastatic (III) model. The ROC area (S.D.) of the 2-tiered model was 0.57 (0.04). Conclusions: ROC analysis optimized the most predictive SEER staging model. The high under staging rate may have prevented patients from selecting definitive radiotherapy after surgery.