• 천문학회지
• /
• 제40권4호
• /
• pp.99-106
• /
• 2007

Statistical analyses were performed to investigate the relative success and accuracy of daily maximum X-ray flux (MXF) predictions, using both multilinear regression and autoregressive time-series prediction methods. As input data for this work, we used 14 solar activity parameters recorded over the prior 2 year period (1989-1990) during the solar maximum of cycle 22. We applied the multilinear regression method to the following three groups: all 14 variables (G1), the 2 so-called 'cause' variables (sunspot complexity and sunspot group area) showing the highest correlations with MXF (G2), and the 2 'effect' variables (previous day MXF and the number of flares stronger than C4 class) showing the highest correlations with MXF (G3). For the advanced three days forecast, we applied the autoregressive timeseries method to the MXF data (GT). We compared the statistical results of these groups for 1991 data, using several statistical measures obtained from a $2{\times}2$ contingency table for forecasted versus observed events. As a result, we found that the statistical results of G1 and G3 are nearly the same each other and the 'effect' variables (G3) are more reliable predictors than the 'cause' variables. It is also found that while the statistical results of GT are a little worse than those of G1 for relatively weak flares, they are comparable to each other for strong flares. In general, all statistical measures show good predictions from all groups, provided that the flares are weaker than about M5 class; stronger flares rapidly become difficult to predict well, which is probably due to statistical inaccuracies arising from their rarity. Our statistical results of all flares except for the X-class flares were confirmed by Yates' $X^2$ statistical significance tests, at the 99% confidence level. Based on our model testing, we recommend a practical strategy for solar X-ray flare predictions.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제13권4호
• /
• pp.2060-2077
• /
• 2019

Recently, mobile healthcare services have attracted significant attention because of the emerging development and supply of diverse wearable devices. Smartwatches and health bands are the most common type of mobile-based wearable devices and their market size is increasing considerably. However, simple value comparisons based on accumulated data have revealed certain problems, such as the standardized nature of health management and the lack of personalized health management service models. The convergence of information technology (IT) and biotechnology (BT) has shifted the medical paradigm from continuous health management and disease prevention to the development of a system that can be used to provide ground-based medical services regardless of the user's location. Moreover, the IT-BT convergence has necessitated the development of lifestyle improvement models and services that utilize big data analysis and machine learning to provide mobile healthcare-based personal health management and disease prevention information. Users' health data, which are specific as they change over time, are collected by different means according to the users' lifestyle and surrounding circumstances. In this paper, we propose a prediction model of user physical activity that uses data characteristics-based long short-term memory (DC-LSTM) recurrent neural networks (RNNs). To provide personalized services, the characteristics and surrounding circumstances of data collectable from mobile host devices were considered in the selection of variables for the model. The data characteristics considered were ease of collection, which represents whether or not variables are collectable, and frequency of occurrence, which represents whether or not changes made to input values constitute significant variables in terms of activity. The variables selected for providing personalized services were activity, weather, temperature, mean daily temperature, humidity, UV, fine dust, asthma and lung disease probability index, skin disease probability index, cadence, travel distance, mean heart rate, and sleep hours. The selected variables were classified according to the data characteristics. To predict activity, an LSTM RNN was built that uses the classified variables as input data and learns the dynamic characteristics of time series data. LSTM RNNs resolve the vanishing gradient problem that occurs in existing RNNs. They are classified into three different types according to data characteristics and constructed through connections among the LSTMs. The constructed neural network learns training data and predicts user activity. To evaluate the proposed model, the root mean square error (RMSE) was used in the performance evaluation of the user physical activity prediction method for which an autoregressive integrated moving average (ARIMA) model, a convolutional neural network (CNN), and an RNN were used. The results show that the proposed DC-LSTM RNN method yields an excellent mean RMSE value of 0.616. The proposed method is used for predicting significant activity considering the surrounding circumstances and user status utilizing the existing standardized activity prediction services. It can also be used to predict user physical activity and provide personalized healthcare based on the data collectable from mobile host devices.

• Journal of Oral Medicine and Pain
• /
• 제26권2호
• /
• pp.133-146
• /
• 2001

This study was performed to predict the conservative treatment outcome of TMD patients by investigating the prognostic factors ; symptom duration, history of previous treatment, history of previous medication, history of trauma, disability of daily activity, severity of pain, noise, limitation of mouth opening(LOM) and maximum comfortable opening(MCO). Two hundreds and fifty-four subjects were selected for this study among the TMD patients who had visited the Dept. of Oral Medicine BNUH and been treated conservatively with medication, physical therapy, behavioral treatment, and splint therapy from 1991 to 2000. The subjects were divided into two groups improved or unimproved according to the treatment response following six months of conservative treatment. Those who showed less than 1 on NAS for pain, TMJ noise, and opening limitation belonged to the improved group and those who showed more than 2 on NAS belonged to the unimproved group. The two groups were compared with respect to symptom severity, number of diagnosis, history of trauma, previous treatment, previous medication, and disability of daily activity. A prognostic equation with the factors revealed to be significantly related to the prognosis of conservative treatment was obtained. The obtained results were as follows ; 1. In improved group, mean duration of history was 12 months, mean treatment duration of a patient was 4 months an mean number of treatment was about 10 times. In other words, in unimproved group, mean duration of history was 27.4 months, mean treatment duration of patient was 10.5 months and mean number of treatment was 19 times. 2. In unimproved group, multiple diagnosis, chronicity, disability of daily activity were significantly greater than that of the improved group. 3. Patients in unimproved group revealed severe noise at first visit and smaller maximum comfortable opening comparatively. 4. Prognostic factors such as duration of treatment, number of treatment, multiplicity, and chronicity and disability of daily activity showed a significant relation in prediction of improvement. 5. Prognostic equation with significant variables is as follows ; Y = 1.984 - 0.251Noise + 0.068MCO - 0.673Multiplicity. - 0.958Chronicity - 0.065Disability. Classification accuracy of 70.3 %, sensitivity of 71.4% and specificity of 66.7% were shown. 6. Prognostic equation with all factors is as follows : Y = 1.599 - 0.038Pain - 0.256Noise - 0.006Limitation + 0.068MCO - 0.580Multiplicity - 1.025Chronicity - 0.720Disability - 0.329Medication - 0.087Treatment + 0.740Trauma. Classification accuracy of 70.3 %, sensitivity of 73% and specificity of 64.3% were shown. 7. Prognostic value of the improved group with significant factors was $1.0446{\pm}1.0726$ and prognostic value of the unimproved group with significant factors was $-0.013{\pm}1.0146$. Prognostic value of the improved group with all factors was $1.0465{\pm}1.0849$ and prognostic value of the unimproved group with all factors was $-0.057{\pm}1.0611$.

• 전기전자학회논문지
• /
• 제14권1호
• /
• pp.17-24
• /
• 2010

본 논문에서는 LBS(Location Based Service) 응용을 위한 독거노인의 일상 활동의 칼로리 소모 예측 모델을 제안한다. 제안한 방식은 PIR (Passive InfraRed) 움직임 센서를 이용하여 독거노인의 활동 패턴을 측정하고, 물리적 활동과 그에 따른 칼로리 소모의 관계를 검토한다. 제안된 움직임 감지 시스템은 개별 노인 주택에 설치되는 센싱 시스템과 중앙 서버 시스템으로 구성된다. 센싱 시스템은 PIR 센서를 장착한 무선 센싱 노드들로 구성된 무선 센서 네트워크 형태로 구현되었고, 각 센서가 감지한 노인들의 움직임은 홈 게이트웨이를 거쳐 중앙 데이터베이스 서버로 저장된다. 서버 시스템은 데이터베이스 서버와 웹 서버로 구성되어 있으며, 웹 기반 모니터링 시스템을 통해 저장된 움직임 데이터를 가공하여 독거노인들의 수발제공자에게 각 노인들의 효율적인 서비스를 제공한다. 실험을 통해 제안한 칼로리 소모 모델의 성능과 적용 가능성을 확인하였다.

• 대한간호학회지
• /
• 제41권4호
• /
• pp.481-490
• /
• 2011

Purpose: The purpose of this study was to identify from the International Classification of Functioning model, factors influencing quality of life in elderly persons and to describe the concrete pathway of influence and the power of each variable. Methods: The sample included 334 elders who lived in 5 districts of D Metropolitan City. A structured questionnaire was used and the collected data were analyzed for fitness, using the AMOS 18.0 program. Results: This model was concise and extensive in predicting the quality of life of elders. The research verified the factors influencing quality of life for elders as direct factors such as activity of daily living (ADL) (${\beta}$=.13, t=2.47), leisure activity (${\beta}$=.55, t=5.04), social disengagement (${\beta}$=-.25, t=-2.25), and depression (${\beta}$=-.62, t=-10.86). Indirect factors including economic status (${\gamma}$=.17, p=.009), type of residence (${\gamma}$=.19, p=.004), ADL (${\gamma}$=.12, p=.027) were important factors in predicting quality of life for elders. These variables explained 75.6% of variance in the prediction model. Conclusion: The findings indicate a need for the nursing scientific community to develop intervention programs considering these variables to improve the quality of life for elders.

• 정보과학회 컴퓨팅의 실제 논문지
• /
• 제21권2호
• /
• pp.154-159
• /
• 2015

과거 어플리케이션 다양성만 지향하던 사용자의 수요가 최근 스마트폰의 고도화된 센서와 기계학습이 결합된 지능형 어플리케이션으로의 선호로 전향되고 있다. 이러한 경향을 반영하여 본 논문에서는 스마트폰에 축적된 사용자의 라이프로깅 데이터에서 의미있는 정보를 추출하고, 추출한 정보를 통해 사용자의 인지적 행동을 대신 가능한 인지 에이전트(Cognitive Agent)개념의 스마트폰-사용자 상호작용 사용자 맞춤 위치-시간-행동 추론 기법을 제안한다. 제안 방법은 사용자의 라이프로깅데이터를 DPGMM (Dirichlet Process Gaussian Mixture Model) 클러스터링 기법으로 사용자 주요 관심지역 POI(Point of Interest)를 자동으로 추출하고, 평생학습이 가능한 강화학습의 한 종류인 POMDP(Partially Observable Markov Decision Process)를 사용하여 사용자의 위치-시간-행동을 추론 한다. 제안 방법으로 구현한 사용자 맞춤 일과 계획 시스템의 시간별 사용자 일과 추론 결과는 70%이상의 성능을 보였으며, 하루 일과 계획 지능형 서비스의 새로운 방향을 제시하고 있다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제7권11호
• /
• pp.2853-2873
• /
• 2013

Over the last few years, one of the most common purposes of smart homes is to provide human centric services in the domain of u-healthcare by analyzing inhabitants' daily living. Currently, the major challenges in activity recognition include the reliability of prediction of each classifier as they differ according to smart homes characteristics. Smart homes indicate variation in terms of performed activities, deployed sensors, environment settings, and inhabitants' characteristics. It is not possible that one classifier always performs better than all the other classifiers for every possible situation. This observation has motivated towards combining multiple classifiers to take advantage of their complementary performance for high accuracy. Therefore, in this paper, a method for activity recognition is proposed by optimizing the output of multiple classifiers with Genetic Algorithm (GA). Our proposed method combines the measurement level output of different classifiers for each activity class to make up the ensemble. For the evaluation of the proposed method, experiments are performed on three real datasets from CASAS smart home. The results show that our method systematically outperforms single classifier and traditional multiclass models. The significant improvement is achieved from 0.82 to 0.90 in the F-measures of recognized activities as compare to existing methods.

• ETRI Journal
• /
• 제45권6호
• /
• pp.1079-1089
• /
• 2023

Stroke is one of the leading causes of long-term disability worldwide, placing huge burdens on individuals and society. Further, automatic human activity recognition is a challenging task that is vital to the future of healthcare and physical therapy. Using a baseline long short-term memory recurrent neural network, this study provides a novel dataset of stretching, upward stretching, flinging motions, hand-to-mouth movements, swiping gestures, and pouring motions for improved model training and testing of stroke-affected patients. A MATLAB application is used to output textual and audible prediction results. A wearable sensor with a triaxial accelerometer is used to collect preprocessed real-time data. The model is trained with features extracted from the actual patient to recognize new actions, and the recognition accuracy provided by multiple datasets is compared based on the same baseline model. When training and testing using the new dataset, the baseline model shows recognition accuracy that is 11% higher than the Activity Daily Living dataset, 22% higher than the Activity Recognition Single Chest-Mounted Accelerometer dataset, and 10% higher than another real-world dataset.

• 한국환경보건학회지
• /
• 제48권3호
• /
• pp.151-158
• /
• 2022

Background: Chemical emissions in the environment have rapidly increased with the accelerated industrialization taking place in recent decades. Residents of industrial complexes are concerned about the health risks posed by chemical exposure. Objectives: This study was performed to suggest modeling methods that take into account multimedia and multi-pathways in human exposure and risk assessment. Methods: The concentration of benzene emitted at industrial complexes in Daesan, South Korea and the exposure of local residents was estimated using the Caltox model. The amount of human exposure based on inhalation rate was stochastically predicted for various activity stages such as resting, normal walking, and fast walking. Results: The coefficient of determination (R²) for the CalTOX model efficiency was 0.9676 and the root-mean-square error (RMSE) was 0.0035, indicating good agreement between predictions and measurements. However, the efficiency index (EI) appeared to be a negative value at -1094.4997. This can be explained as the atmospheric concentration being calculated only from the emissions from industrial facilities in the study area. In the human exposure assessment, the higher the inhalation rate percentile value, the higher the inhalation rate and lifetime average daily dose (LADD) at each activity step. Conclusions: Prediction using the Caltox model might be appropriate for comparing with actual measurements. The LADD of females was higher ratio with an increase in inhalation rate than those of males. This finding would imply that females may be more susceptible to benzene as their inhalation rate increases.

• 정보과학회 컴퓨팅의 실제 논문지
• /
• 제23권2호
• /
• pp.104-109
• /
• 2017

본고에서는 다중 웨어러블 센서 데이터로부터 사용자의 일상 생활 행동 패턴을 분석할 수 있는 새로운 기계학습 모델을 제안한다. 제안하는 모델은 다중 웨어러블 센서 데이터를 효과적으로 학습하기 위하여 사람이 다중 센서 정보를 처리하는 방법을 적용한 새로운 신경망 모델이다. 제안하는 Gated multi-modal neural netoworks는 계층적 신경망 구조를 가지고 있으며 Gate 모듈을 통해 각 센서 데이터를 선택적으로 결합하여 처리하는 특징을 가진다. 실험을 위해 다중 웨어러블 장치를 착용하고 일상 생활 중 한 가지인 레스토랑에서의 행동 센서 데이터를 수집하였다. 실험 결과로서, 제시하는 모델을 이용하여 실제 웨어러블 센서 데이터를 분석하였을 때 분류 정확도가 비교적 정확하고 빠르게 처리할 수 있음을 확인하였다. 또한 모델의 중간 계층에서의 노드의 활성화 패턴 분석을 통해 자동으로 일상생활 패턴을 추출할 수 있고 이를 이용하여 지식 스키마를 생성할 수 있음을 확인하였다.