• 대한원격탐사학회지
• /
• 제40권1호
• /
• pp.57-69
• /
• 2024

Crop classification plays a vitalrole in monitoring agricultural landscapes and enhancing food production. In this study, we explore the effectiveness of Long Short-Term Memory (LSTM) models for crop classification, focusing on distinguishing between apple and rice crops. The aim wasto overcome the challenges associatedwith finding phenology-based classification thresholds by utilizing LSTM to capture the entire Normalized Difference Vegetation Index (NDVI)trend. Our methodology involvestraining the LSTM model using a reference site and applying it to three separate three test sites. Firstly, we generated 25 NDVI imagesfrom the Sentinel-2A data. Aftersegmenting study areas, we calculated the mean NDVI values for each segment. For the reference area, employed a training approach utilizing the NDVI trend line. This trend line served as the basis for training our crop classification model. Following the training phase, we applied the trained model to three separate test sites. The results demonstrated a high overall accuracy of 0.92 and a kappa coefficient of 0.85 for the reference site. The overall accuracies for the test sites were also favorable, ranging from 0.88 to 0.92, indicating successful classification outcomes. We also found that certain phenological metrics can be less effective in crop classification therefore limitations of relying solely on phenological map thresholds and emphasizes the challenges in detecting phenology in real-time, particularly in the early stages of crops. Our study demonstrates the potential of LSTM models in crop classification tasks, showcasing their ability to capture temporal dependencies and analyze timeseriesremote sensing data.While limitations exist in capturing specific phenological events, the integration of alternative approaches holds promise for enhancing classification accuracy. By leveraging advanced techniques and considering the specific challenges of agricultural landscapes, we can continue to refine crop classification models and support agricultural management practices.

• 인터넷정보학회논문지
• /
• 제20권1호
• /
• pp.1-10
• /
• 2019

빅 데이터에서 텍스트 마이닝은 많은 수의 데이터로부터 많은 특징 추출하기 때문에, 클러스터링 및 분류 과정의 계산 복잡도가 높고 분석결과의 신뢰성이 낮아질 수 있다. 특히 텍스트마이닝 과정을 통해 얻는 Term document matrix는 term과 문서간의 특징들을 표현하고 있지만, 희소행렬 형태를 보이게 된다. 본 논문에서는 탐지모델을 위해 텍스트마이닝에서 개선된 GA(Genetic Algorithm)을 이용한 특징 추출 방법을 설계하였다. TF-IDF는 특징 추출에서 문서와 용어간의 관계를 반영하는데 사용된다. 반복과정을 통해 사전에 미리 결정된 만큼의 특징을 선택한다. 또한 탐지모델의 성능 향상을 위해 sparsity score(희소성 점수)를 사용하였다. 스팸메일 세트의 희소성이 높으면 탐지모델의 성능이 낮아져 최적화된 탐지 모델을 찾기가 어렵다. 우리는 fitness function에서 s(F)를 사용하여 희소성이 낮고 TF-IDF 점수가 높은 탐지모델을 찾았다. 또한 제안된 알고리즘을 텍스트 분류 실험에 적용하여 성능을 검증하였다. 결과적으로, 제안한 알고리즘은 공격 메일 분류에서 좋은 성능(속도와 정확도)을 보여주었다.

• 한국전자거래학회지
• /
• 제26권3호
• /
• pp.97-117
• /
• 2021

정기예금 가입 여부 예측은 은행의 대표적인 금융 마케팅 중 하나로, 은행은 다양한 고객 정보를 활용하여 예측 모델을 구성할 수 있다. 정기예금 가입 여부의 분류 정확도를 향상하기 위해, 많은 연구에서 기계학습 기법들을 이용하여 분류 모델들을 개발하였다. 하지만, 이러한 모델들이 만족스러운 성능을 보일지라도 모델의 의사결정 과정에 대한 근거가 적절하게 설명되지 않는다면 산업에서 활용하기가 쉽지 않다. 이러한 문제점을 해결하기 위해, 본 논문은 설명 가능한 정기예금 가입 여부 예측 기법을 제안한다. 먼저, 테이블 형식에서 우수한 성능을 도출하는 의사결정 나무 기반 앙상블 학습 기법인 랜덤 포레스트, GBM, XGBoost, LightGBM을 이용하여 분류 모델들을 개발하고, 10겹 교차검증을 통해 모델들의 분류 성능을 심층 분석한다. 다음으로, 가장 우수한 성능을 도출하는 모델에 설명 가능한 인공지능 기법인 SHAP을 적용하여 고객 정보의 영향도와 의사결정 과정 등을 해석할 수 있는 근거를 제공한다. 제안한 기법의 실용성과 타당성을 입증하기 위해, Kaggle에서 제공한 은행 마케팅 데이터 셋을 대상으로 모의실험을 진행하였으며, 데이터 셋 구성에 따라 GBM과 LightGBM 모델에 SHAP을 각기 적용하여 설명 가능한 정기예금 가입 여부를 위한 분석 및 시각화를 수행하였다.

• 보건행정학회지
• /
• 제4권2호
• /
• pp.27-57
• /
• 1994

The objective of running a long-term care institution is to provide services helpful for maintaining, supporting, and improving elderlies' optimum level of physical, mental, and psychosocial functioning. For the purpose of analyzing the current situations of institutional long term care facilities in Korea, 27 facilities were selected proportionately from each of the cities and provinces, out of the total 152 facilities. About 20% of those who were institutionalized during 25 August through 2 Qctober 1993, the 391 elderlies were chosen on a systematic random basis. The instrument of this study was developed by modifying the tools of CARE, MAI and PCTC. A multivariate approach of discriminant analysis and clustering technique were employed for this study. The Stiudy reveals that there is no clear differentiation of goals and functions among the longterm care institutions in Korea. Staffing patte군 of long-term care facilities shows a shortage of nurses, physical therapists, and dieticians. The linkage between acute care facilities and long-term care is weak, and administration of long-term care faciltiy is carried out by non-professionals. They are responsible for assessing health status before entering the facility, and evaluating elderlies' care. Therefore, it is not surprising to find that most of the facilities have accommodated agede regardless of their real needs and health status. Based upon findings of the analysis, this study has classified long-term care facilities into four types : Type I is to help elderlies maintain independence in daily living activities. Type II facilities have the objective of maintaining and improving the current level of elderlies' function. Type III is to maintain maximum independence of elderlies in activities of daily living. And Type IV is identified for the group of facilities designed to restore or improve functional abilities of elderlies. In conclusion, the following suggestions are made : the need for long-term care should be assessed by multidimensional measurement. Institutional long-term care facilities should be classified and developed in response to type of type of care and service need. Both acute and long-term care facilities should be linked together in order to support the evaluation of service operation and program development.

• 방사성폐기물학회지
• /
• 제14권1호
• /
• pp.63-78
• /
• 2016

To meet nuclear regulatory requirements, more than 95% individual radionuclides in the low- and intermediate-level radioactive waste inventory have to be identified. In this study, the radionuclide inventory has been estimated by taking the long-term radioactive waste generation, the development plan of disposal facility, and the new radioactive waste classification into account. The state of radioactive waste cumulated from 2014 was analyzed for various radioactive sources and future prospects for predicting the long-term radioactive waste generation. The predicted radionuclide inventory results are expected to contribute to secure the development of waste disposal facility and to deploy the safety case for its long-term safety assessment.

• 대한간호학회지
• /
• 제33권2호
• /
• pp.275-283
• /
• 2003

Purpose: This study was to classify elderly in long-term care hospitals for using Resource Utilization Group(RUG-III) and to consider feasibility of payment method based on RUG-III classification system in Korea. Method: This study designed by measuring resident characteristics using the Resident Assessment Instrument-Minimum Data Set(RAI-MDS) and staff time. The data were collected from 382 elderly over sixty-year old, inpatient in the five long-term care hospitals. Staff time was converted into standard time based on the average wage of nurse and aids. Result: The subjects were classified into 4 groups. The group of Clinically Complex was the largest(46.3%), Reduced Physical Function(27.2%), Behavior Problem(17.0%), and Impaired Cognition(9.4%). The average resource use for one resident in terms of care time(nurses, aids) was 183.7 minutes a day. Relative resource use was expressed as a case mix index(CMI) calculated as a proportion of mean resource use. The CMI of Clinically Complex group was the largest(1.10), and then Reduced Physical Function(0.93), Behavior Problem(0.93), and Impaired Cognition(0.83) followed. The difference of the resource use showed statistical significance between major groups(p<0.0001). Conclusion: The results of this study showed that the RUG-III classification system differentiates resources provided to elderly in long-term care hospitals in Korea.

• 대한간호학회지
• /
• 제34권2호
• /
• pp.278-289
• /
• 2004

Purpose: The purpose of this study was to classify the elderly in long-term care facilities using the Resource Utilization Group(RUG-III) and to examine the feasibility of a payment method based on the RUG-III classification system in Korea. Method: This study measured resident characteristics using a Resident Assessment Instrument-Minimum Data Set(RAI-MDS) and staff time. Data was collected from 530 elderly residents over sixty, residing in long-term care facilities. Resource use for individual patients was measured by a wage-weighted sum of staff time and the total time spent with the patient by nurses, aides, and physiotherapists. Result: The subjects were classified into 4 groups out of 7 major groups. The group of Clinically Complex was the largest (46.3%), and then Reduced Physical Function(27.2%), Behavior Problems (17.0%), and Impaired Cognition (9.4%) followed. Homogeneity of the RUG-III groups was examined by total coefficient of variation of resource use. The results showed homogeneity of resource use within RUG-III groups. Also, the difference in resource use among RUG major groups was statistically significant (p<0.001), and it also showed a hierarchy pattern as resource use increases in the same RUG group with an increase of severity levels(ADL). Conclusion: The results of this study showed that the RUG-Ill classification system differentiates resources provided to elderly in long-term care facilities in Korea.

• 한국정보과학회논문지:소프트웨어및응용
• /
• 제26권1호
• /
• pp.167-167
• /
• 1999

The majority of text categorization systems use the term-based classification method. However, because of too many terms, this method is not effective to classify the documents in areal-time environment. This paper presents a real-time concept-based text categorization system,which classifies texts using thesaurus. The system consists of a Korean morphological analyzer, athesaurus tool, and a probability-vector similarity measurer. The thesaurus tool acquires the meaningsof input terms and represents the text with not the term-vector but the concept-vector. Because theconcept-vector consists of semantic units with the small size, it makes the system enable to analyzethe text with real-time. As representing the meanings of the text, the vector supports theconcept-based classification. The probability-vector similarity measurer decides the subject of the textby calculating the vector similarity between the input text and each subject. In the experimentalresults, we show that the proposed system can effectively analyze texts with real-time and do aconcept-based classification. Moreover, the experiment informs that we must expand the thesaurustool for the better system.

• 대한수학회지
• /
• 제59권2호
• /
• pp.217-233
• /
• 2022

Although machine learning shows state-of-the-art performance in a variety of fields, it is short a theoretical understanding of how machine learning works. Recently, theoretical approaches are actively being studied, and there are results for one of them, margin and its distribution. In this paper, especially we focused on the role of margin in the perturbations of inputs and parameters. We show a generalization bound for two cases, a linear model for binary classification and neural networks for multi-classification, when the inputs have normal distributed random noises. The additional generalization term caused by random noises is related to margin and exponentially inversely proportional to the noise level for binary classification. And in neural networks, the additional generalization term depends on (input dimension) × (norms of input and weights). For these results, we used the PAC-Bayesian framework. This paper is considering random noises and margin together, and it will be helpful to a better understanding of model sensitivity and the construction of robust generalization.

• 한국정보통신학회:학술대회논문집
• /
• 한국정보통신학회 2018년도 추계학술대회
• /
• pp.477-479
• /
• 2018

본 연구는 인공신경망의 한 종류인 LSTM(Long Short-Term Memory) 기법을 활용하기 위하여 진행하였다. UTD(University of Texas at Dallas)가 공개한 27종 동작 데이터 중 3축 가속도 및 각속도 데이터를 기본 LSTM 및 Deep Residual Bidir-LSTM 기법에 적용하여 행동을 분류해 보았다.