Earticle

This thesis focuses on uncertainty which begins with Knightian Uncertainty. First we introduce a concept of time-varying uncertainty aversion. We find that uncertainty aversion is increasing before the crash and is resolved after the crash, and tends to move together with S&P 500. Second we present a relationship between uncertainty aversion and business condition. We construct a VECM regression and Granger Causality tests. Using credit spread and term spread as indicators of business conditions, we find some interesting results: (1) Uncertainty aversion has significant positive relationship with credit spreads in United States. (2) Uncertainty aversion has no significant relationship with term-spreads. (3) Uncertainty aversion granger causes both credit spreads and term spreads. This implies that with uncertainty aversion we can explain the credit spread puzzle as well as we can predict future business conditions. If today ’s uncertainty increases, tomorrow’s business condition will be worse, and if today’s uncertainty decreases or is resolved, tomorrow’s business condition will be better.

Along with the spread of digital music and recent growth in the digital music industry, the demands for music recommender are increasing. These days, listeners have increasingly preferred to digital real-time streamlining and downloading to listen to music because it is convenient and affordable for the listeners to do that. We use Bayesian learning through weight of listener’s prefered music site such as Melon, Billboard, Bugs Music, Soribada, and Gini. We reflect most popular current songs across all genres and styles for music recommender system using user profile. It is necessary for us to make the task of preprocessing of clustering the preference with weight of listener’s preferred music site with popular music charts. We evaluated the proposed system on the data set of music sites to measure its performance. We reported some of the experimental result, which is better performance than the previous system.

Naive Bayesian classifiers 네이브 베이지안 분류기는 샘플 데이터로부터 쉽게 구현될 수 있는 강력하고도 많이 사용되는 형식의 분류기다. 그러나 강한 조건부 독립성으로 인하여 효율이 저하되는 분류 결과를 초래한다. 일 반적으로 네이브 베이지안 분류기는 연속성을 가진 특징 데이터의 우도를 처리하기 위해 가우시안 분산을 사용한다. 속성들의 확률밀도는 항상 가우시안 분산에 적합한 것만은 아니다. 또 다른 형식의 분류기는 지도학습을 통해 퍼지 규칙과 퍼지집합을 학습할 수 있는 퍼지신경망이다. 퍼지신경망과 네이브 베이지안 분류기간에는 구조적 유사성을 가지고 있기 때문에 퍼지신경망으로 학습된 분산 그래프를 네이브 베이지안 분류기에 적용하고자 하는 방안이 본 연 구의 목적이다. 따라서 네이브 베이지안 분류기에 가우시안 분산 그래프를 사용한 결과와 퍼지 분산 그래프를 사용 한 결과를 비교하였다. 이를 위해 leukemia와 colon의 DNA 마이크로어레이 데이터를 적용하여 분류하였다. 네이브 베이지안 분류기에 퍼지 분산 그래프를 사용한 결과 가우시안 분산 그래프를 사용한 결과보다 더 신뢰성이 있음을 보여주었다.

Naive Bayesian classifiers are a powerful and well-known type of classifiers that can be easily induced from a dataset of sample cases. However, the strong conditional independence assumptions can sometimes lead to weak classification performance. Normally, naive Bayesian classifiers use Gaussian distributions to handle continuous attributes and to represent the likelihood of the features conditioned on the classes. The probability density of attributes, however, is not always well fitted by a Gaussian distribution. Another eminent type of classifier is the neuro-fuzzy classifier, which can learn fuzzy rules and fuzzy sets using supervised learning. Since there are specific structural similarities between a neuro-fuzzy classifier and a naive Bayesian classifier, the purpose of this study is to apply learning distribution graphs constructed by a neuro-fuzzy network to naive Bayesian classifiers. We compare the Gaussian distribution graphs with the fuzzy distribution graphs for the naive Bayesian classifier. We applied these two types of distribution graphs to classify leukemia and colon DNA microarray data sets. The results demonstrate that a naive Bayesian classifier with fuzzy distribution graphs is more reliable than that with Gaussian distribution graphs.

1인 창조기업에 대한 정부의 지원이 확대되고, 인공지능과 IT 기술의 발달로 1인 창조기업의 성장할 수 있는 환경이 좋아졌다. 이로 인해 1인 창조기업에 대한 사람들의 관심이 증가하고, 1인 창조기업 지원센터를 찾는 청년 창업가의 숫자도 크게 증가하고 있다. 본 연구는 비지도 기계학습과 베이지안 네트워크를 결합한 새로운 유형의 확률적 구조방정식 모형, 즉 PSEM을 사용하여 국내 1인 창조기업의 당기순이익에 영향을 주는 요인을 분석하고자 한다. 아울러 해당 영향요인들이 국내 1인 창조기업의 당기순이익에 미치는 인과관계 값을 산출하기 위하여 What-If 분석을 실시하였다. 분석 결과 당기 순이익의 변화에 큰 영향을 받지 않는 요인들은 서비스 업종일 때 ‘초기 자본금’, ‘창업자특성’, ‘준비 기간’, ‘창업자 배경’, ‘회사 특성’으로 나타났다. 서비스 업종을 제외하였을 때는 ‘창업자 배경’, ‘인식’, ‘초기 자본금’, ‘회사 특성’, ‘창업자특성’으로 분석되었다. 당기순이익에 영향을 주는 요인은 ‘수익’, ‘종업원’, ‘판매’임을 발견하였다.

In the era of economic recession caused by covid-19 pandemic, government expanded various types of supports for one-man creative companies with a bid to promoting healthier ecosystem for start-ups. Furthermore, the 4th industrial revolution atmospheres emboldened by AI technologies have countenanced young entrepreneurs’attempts to launch their own one-man creative companies. However, despite positive trends with one-man creative companies like this, few studies exist investigating which factors affect its net profit. To fill the research void, this study proposes a novel method of probabilistic structural equation model (PSEM) combining unsupervised machine learning and Bayesian networks. Empirical results revealed that those influential factors affecting one-man creative companies’ net profit encompass ‘income’, ‘number of employees’, and ‘sales’. On the contrary, multiple factors initially deemed to be significant to the net profit were excluded, like “background of the CEO”, “characteristics of the company”, “initial fund”, ‘preparation’, and “characteristics of the CEO” in service company. The others company’s results showed that ‘background of the CEO’, ‘perception’, ‘initial fund’, ‘characteristics of the company’, and ‘characteristics of the CEO’ were not included in factors affected by changes of net profit.

The English article system has been known one of the most challenging components of English grammar for foreign learners to master, which needs continuous learning and feedback through assessments. However, current cognitive diagnostic models(CDMs) show some serious limitations: heavy reliance on large scale data, inability to model skill hierarchies and insufficient flexibility to efficiently handle small but repeated measurements. The purpose of the current study was to examine the potential of Bayesian network-based cognitive diagnostic modeling(BN-CDM) as an alternative to the current CDMs. A group of 124 college students(98 females and 26 males) joined a weekly 10-min learning session of the English article system throughout a semester and were administered a series of diagnostic tests. The test data were analyzed using conventional CDMs and BN-CDM. The results show that BN-CDM can handle small but repeated test data much more efficiently than conventional CDMs with a full consideration of hierarchical structures of the subject domain. The study also discusses some pedagogical implications of the results.

In order to demonstrate the effectiveness of sparse representation techniques for speaker recognition, a dictionary of feature vectors belonging to all speakers is constructed by total variability i-vectors. Each feature vector from unknown utterance is expressed as linear weighted sum of a dictionary. The weights are calculated using Block Sparse Bayesian Learning (BSBL) where the sparsest solution can be obtained. By exploiting the speech signal’s block structure and intra-block correlation, the system performance is improved. The experimental results validate that our method outperforms the baseline systems and the system using Orthogonal Matching Pursuit (OMP) algorithm on the typical corpus and realizes the identity validation function.

We explore the software metrics and build a Bayesian Network Model for defect prediction. Much previous work has concentrated on how to select the software metrics that are most likely to indicate fault-proneness, based on the hypnosis that these metrics are independent. But in reality, software metric values are predicted not only correlated with fault-proneness, but also observed internal complex relationship with each other. In this paper, we build a Bayesian network model to represent the probability distribution of each factor and how they affect defects, considering strong or weak correlations are existed between individual metric attributes. We perform a comparative experimental study of effectiveness of Bayesian Network, logistic regression and Naive Bayes on a public data set from an open source software system. The result shows that our approach produces statistically significant estimations.

Bayesian network is a directed acyclic graph. Existing Bayesian network learning approaches based on search & scoring usually work with a heuristic search for finding the highest scoring structure. This paper describes a new data mining algorithm to learn Bayesian networks structures based on an immune binary particle swarm optimization (IB-PSO) method and the Minimum Description Length (MDL) principle. IBPSO is proposed by combining the immune theory in biology with particle swarm optimization (PSO). It constructs an immune operator accomplished by two steps, vaccination and immune selection. The purpose of adding immune operator is to prevent and overcome premature convergence. Experiments show that IBPSO not only improves the quality of the solutions, but also reduces the time cost.

Aiming at the cuckoo search algorithm (CSA) with disadvantages of slow convergence speed, getting into local extremum easily and low accuracy, we put forward cuckoo search algorithm with cauchy mutation(CCSA). For Bayesian networks(BNs) structure learning with insufficient data, we propose data completion method and Bayesian network structure learning with insufficient data based on CCSA(BNSL-ID-CCSA). In BNSL-ID-CCSA, firstly, we adopt K2 metric as evaluation measure for learning Bayesian networks from data. Secondly, we use expectation maximization(EM) algorithm and CCSA to make BNSL-ID-CCSA quickly and accurately converge to the global optimal solution. The experimental results show that BNSL-ID-CCSA has strong learning ability and good stability.

In this paper, we present a new method for automatic image annotation by applying semi- supervised learning based on the Bayesian framework. On the one hand, we employ the semi- supervised learning, i.e., transductive support vector machine (TSVM) to enhance the quality of training image data, which is a promising way to find out the underlying relevant data from the unlabeled ones. On the other hand, a simple yet very efficient Bayesian model is built to implement image annotation by the maximum a posteriori (MAP) criterion. The novelty of our method mainly lies in two aspects: exploiting TSVM to improve the quality of training image dataset and utilizing the Bayesian model to predict the candidate annotations for the unseen images. Experimental results on the standard Corel dataset demonstrate that the proposed method is superior or highly competitive to several state-of-the-art approaches.

베이지안망은 불확실한 상황 하에서 영역지식을 표현하고 예측하기 위한 좋은 도구로 알려져 있다. 그러나 변수가 많아졌을 때 학습이 어렵고 시간의 요구량이 늘어나게 되어 효율적이고 신뢰도 높은 탐색에 문제가 있다. 이를 해결하기 위해서 노드의 순서를 정하여 효율적인 구조학습이 가능하도록 한다. 본 논문에서는 각 상황에 따른 확률의 엔트로피를 계산하여 다양한 변수간의 관계나 상호의존적인 상황에서도 오차를 줄이고 신뢰도를 높일 수 있는 효과적인 분류학습모델을 제시한다. 베이지안망 학습 방법 중 일반적으로 널리 알려져 있는 K2알고리즘에서 각 노드의 엔트로피 수치를 계산하여 엔트로피가 낮은 노드의 순서를 결정하여 결과적으로 빠른 시간 안에 최적화된 베이지안망의 모델을 구성하는 효율적인 학습모델을 제시한다.

Bayesian networks are known as the best tools to express and predict the domain knowledge with uncertain environments. However, bayesian learning could be too difficult to do effective and reliable searching. To solve the problems of overtime demand, the nodes should be arranged orderly, so that effective structural learning can be possible. This paper suggests the classification learning model to reduce the errors in the independent condition, in which a lot of variables exist and data can increase the reliability by calculating the each entropy of probabilities depending on each circumstances. Also efficient learning models are suggested to decide the order of nodes, that has lowest entropy by calculating the numerical values of entropy of each node in K2 algorithm. Consequently the model of the most suitably settled Bayesian networks could be constructed as quickly as possible.

This paper studies optimal consumption and investment behaviors of an individual when risky asset returns and her income are affected by the business cycle. The investor considers the incomplete information risk of unobservable macroeconomic conditions and updates her belief of expected risky asset returns through Bayesian learning. We find that the optimal investment strategy, certainty equivalent wealth, and portfolio hedging demand significantly depend on the belief about the macroeconomic conditions.

Along with the spread of digital music and recent growth in the digital music industry, the demands for music recommender are increasing. These days, listeners have increasingly preferred to digital real-time streamlining and downloading to listen to music because it is convenient and affordable for the listeners to do that. We use Bayesian learning through weight of listener's prefered music site such as Melon, Billboard, Bugs Music, Soribada, and Gini. We reflect most popular current songs across all genres and styles for music recommender system using user profile. It is necessary for us to make the task of preprocessing of clustering the preference with weight of listener's preferred music site with popular music charts. We evaluated the proposed system on the data set of music sites to measure its performance. We reported some of the experimental result, which is better performance than the previous system.

In this paper, we will evaluate the power and usefulness of Bayesian network classifiers for financial credit scoring. Various types of Bayesian network classifiers will be evaluated and contrasted including unrestricted Bayesian network classifiers learnt using a global score metric. For comparison, two of statistical classifiers will be evaluated. The experiments will be carried out on three real life credit scoring data sets. It will be shown that general Bayesian network classifiers learned by a global score search have a good performance and by using the Markov blanket concept, a natural form of input selection is obtained, which results in parsimonious and powerful models for financial credit scoring.

Artificial intelligence (AI) is making computer systems intelligent to do right thing. The AI is used today in a variety of fields, such as journalism, medical, industry as well as entertainment. The impact of AI is becoming larger day after day. In general, the AI system has to lead the optimal decision under uncertainty. But it is difficult for the AI system can derive the best conclusion. In addition, we have a trouble to represent the intelligent capacity of AI in numeric values. Statistics has the ability to quantify the uncertainty by two approaches of frequentist and Bayesian. So in this paper, we propose a methodology of the connection between statistics and AI efficiently. We compute a fixed value for estimating the population parameter using the frequentist learning. Also we find a probability distribution to estimate the parameter of conceptual population using Bayesian learning. To show how our proposed research could be applied to practical domain, we collect the patent big data related to Apple company, and we make the AI more intelligent to understand Apple's technology.

Nowadays, due to development of automatic control devices and various sensors, one operator can freely handle several remote plants and processes. Automatic diagnosis and warning systems have been adopted in various fields, in order to prepare an operator's absence for patrolling plants. In this paper, a Bayesian networks based on-line diagnosis system is proposed for a wastewater treatment process. Especially, the suggested system is included learning structure, which can continuosly update conditional probabilities in the networks. To evaluate performance of proposed model, we made a lab-scale five-stage step-feed enhanced biological phosphorous removal process plant and applied on-line diagnosis system to this plant in the summer.

A new structure learning approach for Bayesian networks based on asexual reproduction optimization (ARO) is proposed in this paper. ARO can be considered an evolutionary-based algorithm that mathematically models the budding mechanism of asexual reproduction. In ARO, a parent produces a bud through a reproduction operator; thereafter, the parent and its bud compete to survive according to a performance index obtained from the underlying objective function of the optimization problem: This leads to the fitter individual. The convergence measure of ARO is analyzed. The proposed method is applied to real-world and benchmark applications, while its effectiveness is demonstrated through computer simulations. Results of simulations show that ARO outperforms genetic algorithm (GA) because ARO results in a good structure and fast convergence rate in comparison with GA.