On Uncertain Probabilistic Data Modeling

Teng Lv,Ping Yan,Weimin He 보안공학연구지원센터 2016 International Journal of Database Theory and Appli Vol.9 No.12

Uncertainty in data is caused by various reasons including data itself, data mapping, and data policy. For data itself, data are uncertain because of various reasons. For example, data from a sensor network, Internet of Things or Radio Frequency Identification is often inaccurate and uncertain because of devices or environmental factors. For data mapping, integrated data from various heterogonous data sources is commonly uncertain because of uncertain data mapping, data inconsistency, missing data, and dirty data. For data policy, data is modified or hided for policies of data privacy and data confidentiality in an organization. But traditional deterministic data management mainly deals with deterministic data which is precise and certain, and cannot process uncertain data. Modeling uncertain data is a foundation of other technologies for further processing data, such as indexing, querying, searching, mapping, integrating, and mining data, etc. Probabilistic data models of relational databases, XML data and graph data are widely used in many applications and areas today, such as World Wide Web, semantic web, sensor networks, Internet of Things, mobile ad-hoc networks, social networks, traffic networks, biological networks, genome databases, and medical records, etc. This paper presents a survey study of different probabilistic models of uncertain data in relational databases, XML data, and graph data, respectively. The advantages and disadvantages of each kind of probabilistic modes are analyzed and compared. Further open topics of modeling uncertain probabilistic data such as semantic and computation aspects are discussed in the paper. Criteria for modeling uncertain data, such as expressive power, complexity, efficiency, extension are also proposed in the paper.

확률적 머신러닝 모델기반의 리튬이온배터리 파라미터 추정 알고리즘

김민호(Minho Kim),송민석(Minseok Song),임정택(Jeongtaek Lim),함경선(Kyung Sun Ham),이도헌(DOHEON LEE),김태형(Taehyoung Kim) 한국에너지학회 2024 에너지공학 Vol.33 No.1

In this study, a new lithium-ion battery performance degradation model and a stochastic machine learning model-based lithium-ion battery parameter estimation method were proposed and verified through actual battery degradation cycle experiment data. The proposed parameter estimation method based on a stochastic machine learning model requires less battery model operation time compared to other methods, enabling efficient parameter estimation. The lithium-ion battery performance degradation model is an equivalent circuit-based model, but it reflects various electrochemical phenomena, including side reactions on the surface of the anode active material, including the formation of a solid electrolyte interphase (SEI) layer, the loss of positive electrode active material due to mechanical stress-induced fatigue failure is included, and the corresponding decrease in the amount of cyclable lithium. In the proposed method of estimating the parameters of a lithium-ion battery model, a probabilistic machine learning model that can estimate battery model parameters from sensible data such as voltage and current is developed and used to generate virtual experiment data. We proposed a technique for learning and finding optimal battery model parameters based on the learned model. The developed performance degradation model and parameter estimation method were verified based on actual experimental data. Since it is impossible to observe the inside of the battery, correct answers to the battery parameters cannot be obtained, so the model and parameter estimation algorithm are indirectly verified through errors of voltage and temperature. As a result of the verification, the errors in voltage and temperature were found to be 0.676% and 0.207%, respectively.

이산화 과정을 배제한 실수 값 인자 데이터의 고차 패턴 분석을 위한 진화연산 기반 하이퍼네트워크 모델 (pp.120-128)

하정우(Jung-Woo Ha),장병탁(Byoung-Tak Zhang) 한국정보과학회 2010 정보과학회논문지 : 소프트웨어 및 응용 Vol.37 No.2

하이퍼네트워크는 하이퍼그래프의 일반화된 모델로 학습과정에 있어 진화적 개념을 도입한 확률 그래프 기반의 기계학습 알고리즘으로서 최근 들어 여러 다양한 분야에 응용되고 있다. 그러나 하이퍼 네트워크 모델은 데이터와 모델을 구성하는 하이퍼에지 간의 동등비교를 기반으로 하는 학습과정의 특성상 데이터를 구성하는 인자들이 범주형인 경우에만 학습 및 모델링이 가능하고 실수 값으로 표현된 데이터를 학습하기 위해서는 이산화 등의 전처리가 선행되어야 한다는 한계점이 있다. 하지만 데이터 전처리에 있어 이산화 하는 과정은 필연적으로 정보손실이 발생할 수밖에 없기 때문에 이는 분류 예측 모델의 성능저하를 유발하는 원인이 될 수 있다. 이러한 기존 하이퍼네트워크 모델의 한계점을 극복하기 위해 본 연구에서는 별도의 데이터 전처리 과정을 거치지 않고 실수 인자로 구성된 데이터의 패턴 학습이 가능한 개선된 하이퍼네트워크 모델을 제안한다. 여러 실험 결과를 통해 제안한 하이퍼네트워크 모델은 기존 하이퍼 네트워크 모델에 비해 실수형 데이터에 대한 학습 및 분류 결과 성능이 향상되었을 뿐 아니라, 다른 여러기계학습 방법들에 비해서도 경쟁력 있는 성능이 나타남을 확인하였다. A hypernetwork is a generalized hypergraph and a probabilistic graphical model based on evolutionary learning. Hypernetwork models have been applied to various domains including pattern recognition and bioinformatics. Nevertheless, conventional hypernetwork models have the limitation that they can manage data with categorical or discrete attibutes only since the learning method of hypernetworks is based on equality comparison of hyperedges with learned data. Therefore, realvalued data need to be discretized by preprocessing before learning with hypernetworks. However, discretization causes inevitable information loss and possible decrease of accuracy in pattern classification. To overcome this weakness, we propose a novel feature-wise L1-distance based method for real-valued attributes in learning hypernetwork models in this study. We show that the proposed model improves the classification accuracy compared with conventional hypernetworks and it shows competitive performance over other machine learning methods.

Probabilistic Lifetime Prediction of Electronic Packages Using Advanced Uncertainty Propagation Analysis and Model Calibration

Hyunseok Oh,Hsiu-Ping Wei,Bongtae Han,Youn, Byeng D. IEEE 2016 IEEE transactions on components, packaging, and ma Vol.6 No.2

<P>We propose a novel methodology for calibrating the physics-based lifetime models of the electronic packages using the eigenvector dimension-reduction (EDR) method and a censored data analysis. The methodology enables to overcome two challenges that are encountered in typical electronic packaging applications: 1) the minimum computational cost without sacrificing the prediction accuracy and 2) the proper handling of the censored data. The EDR method is first employed for uncertainty propagation for the computational efficiency when multiple unknown variables are to be used in nonlinear damage models. Next, the likelihood function is modified to handle the failure data as well as the censored data in the likelihood analysis, and thus establishes the correlation between the model response and the experimental result. Finally, through an unconstrained optimization process, a calibrated parameter set of statistical distributions for unknown input variables is obtained while maximizing the modified likelihood. The proposed statistical calibration approach is implemented for solder joint fatigue reliability. The results confirm the claimed computational effectiveness for an accurate physics-based lifetime model.</P>

The Development of Probabilistic Time and Cost Data: Focus on field conditions and labor productivity

Hyun, Chang-Taek,Hong, Tae-Hoon,Ji, Soung-Min,Yu, Jun-Hyeok,An, Soo-Bae Korea Institute of Construction Engineering and Ma 2011 Journal of construction engineering and project ma Vol.1 No.1

Labor productivity is a significant factor associated with controlling time, cost, and quality. Many researchers have developed models to define methods of measuring the relationship between productivity and various parameters such as the size of working area, maximum working hours, and the crew composition. Most of the previous research has focused on estimating productivity; however, this research concentrates on estimating labor productivity and developing time and cost data for repetitive concrete pouring activity. In Korea, "Standard Estimating" only entails the average productivity data of the construction industry, and it is difficult to predict the time and cost spent on any particular project. As a result, errors occur in estimating duration and cost for individual activities or projects. To address these issues, this research sought to collect data, measure productivity, and develop time and cost data using labor productivity based on field conditions from the collected data. A probabilistic approach is also proposed to develop data. A case study is performed to validate this process using actual data collected from construction sites. It is possible that the result will be used as the EVMS baseline of cost management and schedule management.

무선 센서 네트워크에서 동적 클러스터 유지 관리 방법을 이용한 에너지 효율적인 주기적 데이터 수집

윤상훈,조행래,Yun, SangHun,Cho, Haengrae 대한임베디드공학회 2010 대한임베디드공학회논문지 Vol.5 No.4

Wireless sensor networks (WSNs) are used to collect various data in environment monitoring applications. A spatial clustering may reduce energy consumption of data collection by partitioning the WSN into a set of spatial clusters with similar sensing data. For each cluster, only a few sensor nodes (samplers) report their sensing data to a base station (BS). The BS may predict the missed data of non-samplers using the spatial correlations between sensor nodes. ASAP is a representative data collection algorithm using the spatial clustering. It periodically reconstructs the entire network into new clusters to accommodate to the change of spatial correlations, which results in high message overhead. In this paper, we propose a new data collection algorithm, name EPDC (Energy-efficient Periodic Data Collection). Unlike ASAP, EPDC identifies a specific cluster consisting of many dissimilar sensor nodes. Then it reconstructs only the cluster into subclusters each of which includes strongly correlated sensor nodes. EPDC also tries to reduce the message overhead by incorporating a judicious probabilistic model transfer method. We evaluate the performance of EPDC and ASAP using a simulation model. The experiment results show that the performance improvement of EPDC is up to 84% compared to ASAP.

건축물 에너지 수요 예측을 위한 이종 공공 데이터 통합 방안 연구

김철호(Chulho Kim),김한주(Hanjoo Kim),변지욱(Jiwook Byun),고재현(Jaehyun Go),허연숙(Yeonsook Heo) 대한설비공학회 2022 대한설비공학회 학술발표대회논문집 Vol.2022 No.6

This study investigated a method of integrating building energy-consumption public data such as MOLIT EAIS and KEPCO AMI to develop the probabilistic models to predict the electricity demand of non-residential buildings. The data set was created by integrating electricity data composed by consideration of time (24 hours) of 9 business types into 5 building types (i.e., education, office, hotel, culture, retail). Individual building energy data source alone provides partial information, and different data sources are at different temporal resolution. Therefore, there is a strong need to develop a framework to integrate various types of public data sets, and this data-integration framework will be essential to develop building energy forecasting models at high resolution levels.

건축물 에너지 수요 예측을 위한 이종 공공 데이터 통합 방안 연구

김철호(Chulho Kim),김한주(Hanjoo Kim),변지욱(Jiwook Byun),고재현(Jaehyun Go),허연숙(Yeonsook Heo) 대한설비공학회 2022 대한설비공학회 학술발표대회논문집 Vol.2022 No.6

This study investigated a method of integrating building energy-consumption public data such as MOLIT EAIS and KEPCO AMI to develop the probabilistic models to predict the electricity demand of non-residential buildings. The data set was created by integrating electricity data composed by consideration of time (24 hours) of 9 business types into 5 building types (i.e., education, office, hotel, culture, retail). Individual building energy data source alone provides partial information, and different data sources are at different temporal resolution. Therefore, there is a strong need to develop a framework to integrate various types of public data sets, and this data-integration framework will be essential to develop building energy forecasting models at high resolution levels.

건축물 에너지 수요 예측을 위한 이종 공공 데이터 통합 방안 연구

김철호(Chulho Kim),김한주(Hanjoo Kim),변지욱(Jiwook Byun),고재현(Jaehyun Go),허연숙(Yeonsook Heo) 대한설비공학회 2022 대한설비공학회 학술발표대회논문집 Vol.2022 No.6

This study investigated a method of integrating building energy-consumption public data such as MOLIT EAIS and KEPCO AMI to develop the probabilistic models to predict the electricity demand of non-residential buildings. The data set was created by integrating electricity data composed by consideration of time (24 hours) of 9 business types into 5 building types (i.e., education, office, hotel, culture, retail). Individual building energy data source alone provides partial information, and different data sources are at different temporal resolution. Therefore, there is a strong need to develop a framework to integrate various types of public data sets, and this data-integration framework will be essential to develop building energy forecasting models at high resolution levels.

A Multi-target Tracking Algorithm for Application to Adaptive Cruise Control

Moon Il-ki,Yi Kyongsu,Cavency Derek,Hedrick J. Karl The Korean Society of Mechanical Engineers 2005 JOURNAL OF MECHANICAL SCIENCE AND TECHNOLOGY Vol.19 No.9

This paper presents a Multiple Target Tracking (MTT) Adaptive Cruise Control (ACC) system which consists of three parts; a multi-model-based multi-target state estimator, a primary vehicular target determination algorithm, and a single-target adaptive cruise control algorithm. Three motion models, which are validated using simulated and experimental data, are adopted to distinguish large lateral motions from longitudinally excited motions. The improvement in the state estimation performance when using three models is verified in target tracking simulations. However, the performance and safety benefits of a multi-model-based MTT-ACC system is investigated via simulations using real driving radar sensor data. The MTT-ACC system is tested under lane changing situations to examine how much the system performance is improved when multiple models are incorporated. Simulation results show system response that is more realistic and reflective of actual human driving behavior.