• The KIPS Transactions:PartB
• /
• v.13B no.7 s.110
• /
• pp.643-652
• /
• 2006

Decision trees are mainly used for the classification and prediction in data mining. The distribution of spatial data and relationships with their neighborhoods are very important when conducting classification for spatial data mining in the real world. Spatial decision trees in previous works have been designed for reflecting spatial data characteristic by rating Euclidean distance. But it only explains the distance of objects in spatial dimension so that it is hard to represent the distribution of spatial data and their relationships. This paper proposes a decision tree based on spatial entropy that represents the distribution of spatial data with the dispersion and dissimilarity. The dispersion presents the distribution of spatial objects within the belonged class. And dissimilarity indicates the distribution and its relationship with other classes. The rate of dispersion by dissimilarity presents that how related spatial distribution and classified data with non-spatial attributes we. Our experiment evaluates accuracy and building time of a decision tree as compared to previous methods. We achieve an improvement in performance by about 18%, 11%, respectively.

• Nuclear Engineering and Technology
• /
• v.53 no.8
• /
• pp.2534-2546
• /
• 2021

Nuclear emergency preparedness and response is an essential part to ensure the safety of nuclear power plant (NPP). Key support technologies of nuclear emergency decision-making usually consist of accident diagnosis, source term estimation, accident consequence assessment, and protective action recommendation. Source term estimation is almost the most difficult part among them. For example, bad communication, incomplete information, as well as complicated accident scenario make it hard to determine the reactor status and estimate the source term timely in the Fukushima accident. Subsequently, it leads to the hard decision on how to take appropriate emergency response actions. Hence, this paper aims to develop a method for rapid source term estimation to support nuclear emergency decision making in pressurized water reactor NPP. The method aims to make our knowledge on NPP provide better support nuclear emergency. Firstly, this paper studies how to build a Bayesian network model for the NPP based on professional knowledge and engineering knowledge. This paper presents a method transforming the PRA model (event trees and fault trees) into a corresponding Bayesian network model. To solve the problem that some physical phenomena which are modeled as pivotal events in level 2 PRA, cannot find sensors associated directly with their occurrence, a weighted assignment approach based on expert assessment is proposed in this paper. Secondly, the monitoring data of NPP are provided to the Bayesian network model, the real-time status of pivotal events and initiating events can be determined based on the junction tree algorithm. Thirdly, since PRA knowledge can link the accident sequences to the possible release categories, the proposed method is capable to find the most likely release category for the candidate accidents scenarios, namely the source term. The probabilities of possible accident sequences and the source term are calculated. Finally, the prototype software is checked against several sets of accident scenario data which are generated by the simulator of AP1000-NPP, including large loss of coolant accident, loss of main feedwater, main steam line break, and steam generator tube rupture. The results show that the proposed method for rapid source term estimation under nuclear emergency decision making is promising.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.32B no.8
• /
• pp.1045-1056
• /
• 1995

The purpose of multiple alarm processing is to give the operator the correct information and perception of the malfunction present in the plant. In this thesis, an APS(Alarm Processing System) is studied for fossil power plants. This APS is based on a cause-consequence trees in the knowledge representation aspect for alarm and plant and adapts alarm filtering methods using fired time information in the decision aspect. Through the cause-consequence trees and filtering methods, the Alarm Processing System finds the cause alarm among the fired multiple alarms and calculates the cause degree which represents the possibility of a fault occurring in the instruments of the plant with the information of fired alarm. The knowledge base is built via interviews and questionaries with the expert operators on the Seoul power plant unit 4. Finally, the validity of the studied APS is shown via simulations.

• KIEE International Transactions on Electrophysics and Applications
• /
• v.11C no.2
• /
• pp.12-17
• /
• 2001

In this work, the correlation between the PD Patterns and the electrical tree propagation has been investigated by use of the specimen removed from the insulation of the real 154 kV XLPE underground power cables. As a result, it could be deduced that the PD pattern regarding electrical trees depends on their type which could be classified into three different distinct groups such as branch-bush mixed. Considering the results of our investigation, if the partial discharge magnitude is only considered for the diagnosis of the cable system, it is possible to draw a wrong decision. Therefore, it is possible to propose that the time characteristics of PD pattern should be taken into account for the diagnosis of the cable system in addition to the conventional $\Phi$-q-n characteristics.

• Management Science and Financial Engineering
• /
• v.19 no.1
• /
• pp.1-10
• /
• 2013

Hierarchical forecasting strategy does not always outperform direct forecasting strategy. The performance generally depends on demand features. This research guides the use of the alternative forecasting strategies according to demand features. This paper developed and evaluated various classification models such as logistic regression (LR), artificial neural networks (ANN), decision trees (DT), boosted trees (BT), and random forests (RF) for predicting the relative performance of the alternative forecasting strategies for the South Korean navy's spare parts demand which has non-normal characteristics. ANN minimized classification errors and inventory costs, whereas LR minimized the Brier scores and the sum of forecasting errors.

• Journal of Korea Society of Digital Industry and Information Management
• /
• v.20 no.2
• /
• pp.1-10
• /
• 2024

For the purpose of predicting credit card customer churn accurately through data analysis, a model can be constructed with various machine learning algorithms, including decision tree. And feature importance has been utilized in selecting better input features that can improve performance of data analysis models for several application areas. In this paper, a method of utilizing feature importance calculated from the MDI method and its effects are investigated in the credit card customer churn prediction problem with classification trees. Compared with several random feature selections from case data, a set of input features selected from higher value of feature importance shows higher predictive power. It can be an efficient method for classifying and choosing input features necessary for improving prediction performance. The method organized in this paper can be an alternative to the selection of input features using feature importance in composing and using classification trees, including credit card customer churn prediction.

• Korean Management Science Review
• /
• v.21 no.3
• /
• pp.101-113
• /
• 2004

In the rapidly propagating Internet based electronic transaction environment. the importance of real time order confirmation has been more emphasized, In this paper, using data mining techniques, we develop intelligent operations decision model to allow real time order confirmation at the time the customer places an order with required delivery terms. Among various operation plannings used for order fulfillment. mill routing is the first interface decision point to link the order receiving at the marketing with the production planning for order fulfillment. Though linear programming based mathematical optimization techniques are mostly used for mill routing problems, some early orders should wait until sufficient orders are gathered for optimization. And that could effect longer order fulfillment lead-time, and prevent instant order confirmation of delivery terms. To cope with this problem, we provide the intelligent decision model to allow instant order based mill routing decisions. Data mining techniques of decision trees and neural networks. which are more popular in marketing and financial applications, are used to develop the model. Through diverse computational trials with the industrial data from the steel company. we have reported that the performance of the proposed approach is effective compared to the present heuristic only mill routing results. Various issues of data mining techniques application to the mill routing problems having linear programming characteristics are also discussed.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2011.10a
• /
• pp.171-172
• /
• 2011

In this research, we discuss the process of analyzing the solution of four-peg tower of Hanoi using protocol analysis technique, and of developing decision trees from the analyzed contents. For these, we design and implement a computer program that can simulate four-peg tower of Hanoi. The implemented program generates a certain regular-to-regular tower of Hanoi problem, let a user to solve the problem, and records the solution process. From the process by the implemented program and the recorded protocol, we can build the decision tree for the solution. We expect this research can contribute to find a non-optimal solution for n-peg tower of Hanoi.

• Journal of Hospice and Palliative Care
• /
• v.24 no.3
• /
• pp.184-193
• /
• 2021

Purpose: This study attempted to develop clinical guidelines to help patients use hospice and palliative care (HPC) at an appropriate time after writing physician orders for life-sustaining treatment (POLST) by identifying the characteristics of HPC use of patients with terminal cancer. Methods: This retrospective study was conducted to understand the characteristics of HPC use of patients with terminal cancer through decision tree analysis. The participants were 394 terminal cancer patients who were hospitalized at a cancer-specialized hospital in Seoul, South Korea and wrote POLST from January 1, 2019 to March 31, 2021. Results: The predictive model for the characteristics of HPC use showed three main nodes (living together, pain control, and period to death after writing POLST). The decision tree analysis of HPC use by terminal cancer patients showed that the most likely group to use HPC use was terminal cancer patients who had a cohabitant, received pain control, and died 2 months or more after writing a POLST. The probability of HPC usage rate in this group was 87.5%. The next most likely group to use HPC had a cohabitant and received pain control; 64.8% of this group used HPC. Finally, 55.1% of participants who had a cohabitant used HPC, which was a significantly higher proportion than that of participants who did not have a cohabitant (1.7%). Conclusion: This study provides meaningful clinical evidence to help make decisions on HPC use more easily at an appropriate time.

• Korean Journal of Artificial Intelligence
• /
• v.10 no.1
• /
• pp.15-19
• /
• 2022

This study is to understand the relationship between turnover and various conditions. Turnover refers to workers moving from one company to another, which exists in various ways and forms. Currently, a large number of workers are considering many turnover rates to satisfy their income levels, distance between work and residence, and age. In addition, they consider changing jobs a lot depending on the type of work, the decision-making ability of workers, and the level of education. The company needs to accept the conditions required by workers so that competent workers can work for a long time and predict what measures should be taken to convert them into long-term workers. The study was conducted because it was necessary to predict what conditions workers must meet in order to become long-term workers by comparing various conditions and turnover using regression and decision trees. It used Microsoft Azure machines to produce results, and it found that among the various conditions, it looked for different items for long-term work. Various methods were attempted in conducting the research, and among them, suitable algorithms adopted algorithms that classify various kinds of algorithms and derive results, and among them, two decision tree algorithms were used to derive results.