Earticle

오늘날 인터넷이 확산되어감에 따라, e-CRM에 대한 관심이 증대되고 있다. 그 중에서도 특히 '추천시스템'은 e-CRM의 여러 응용분야 중에서도 실무적으로 그리고 학문적으로 가장 활발하게 연구되고 있는 분야 중 하나다. 추천을 위한 여러가지 방법들 중에서, 지금까지 주류를 이뤄온 방법들은 협동 필터링(Collaborative Filtering) 기법과 내용 기반(Content-Based) 접근법이다. 그러나 이러한 기존 방법들은 몇 가지 태생적인 한계점으로 인해 고객의 구매 이력이 많지 않은 중소형 인터넷 쇼핑몰에 적용하기 어렵다는 단점이 있다. 이에, 본 연구에서는 고객의 인구통계 및 구매정보에 2가지 데이터마이닝 기법들(연관 관계 기법과 분류 기법)을 적용하고, 이 결과를 조정 에이전트를 통해 결합하는 형태의 새로운 추천 시스템의 모형과 시스템 구조 체계를 제안한다. 제안된 연구 모형의 유용성을 검증하기 위해, 본 연구에서는 실제 사례에 적용한 웹 기반 프로토타입을 개발, 활용하였다. 프로토타입의 유용성을 실제 사용자들로부터 설문을 통해 조사해 본 결과, 본 연구에서 제안한 추천모형이 생성한 맞춤 정보가 사용자들에게 매우 유익하게 인지됨을 확인하였다.

As the Internet spreads, many people have interests in e-CRM and product recommender systems, one of e-CRM applications. Among various approaches for recommendation, collaborative filtering and content-based approaches have been investigated and applied widely. Despite their popularity, traditional recommendation approaches have some limitations. They require at least one purchase transaction per user. In addition, they don't utilize much information such as demographic and specific personal profile information. This study suggests new hybrid recommendation model using two data mining techniques, association rule and classification, as well as intelligent agent to overcome these limitations. To validate the usefulness of the model, it was applied to the real case and the prototype web site was developed. We assessed the usefulness of the suggested recommendation model through online survey. The result of the survey showed that the information of the recommendation was generally useful to the survey participants.

Associative classification is a supervised classification approach, integrating association mining and classification. Several studies in data mining have shown that associative classification achieves higher classification accuracy than do traditional classification techniques. However, the associative classification suffers from a major drawback: The huge number of the generated classification rules which takes efforts to select the best ones in order to construct the classifier. To overcome such drawback, we have proposed an associative classification method that reduces associative classification rules without jeopardizing the classification accuracy. Moreover, we will introduce in this paper two different strategies to classify new instances based on some interestingness measures that arise from data mining literature in order to select the best rules during classification. A detailed description of this method is presented in this paper, as well as the experimentation study on 12 benchmark data sets proving that our approach is highly competitive in terms of accuracy in comparison with popular classification approaches.

To produce a new Rule, many rule-based classifiers use one measure to select the best attribute-value pair. So, a lot of attribute-values have the same best values, and we cannot distinguish which attribute-value pair is the best. On the other hand, these classifiers usually combine the best attribute-value pairs to produce rules, whether they bias toward the same class label or not. To address issues, this paper proposes a new measure approach named deviation. Using the attribute-value deviation, it is easy to distinguish which attribute-value pairs bias toward the same class label. In this paper, we propose a multi-measure called effect measure to select the best attribute-value pair. It integrates the deviation and the entropy, and we also propose a new classification approach called CAEM which uses the effect measure to select the best attribute-value pairs. Experimental results show the method of multi-measure is necessary.

This paper introduces a very important classification aspect for the analysis of huge amount of data stored in databases and other repositories. Numerous classification models are available in the literature, to predict the class of objects whose class level is unknown. Literature reveals that most of the available models are not capable in handling imperfect data. In view of this, present paper proposes a new rough set based classification model to derive the classification (IF-THEN) rules. Furthermore, developed model has been applied to handle bank-loan applications database as either safe, unsafe or risky. However, proposed model can also be used for the analysis of data from other domains.

This paper is focused on how to utilize Change in Tariff Classification (CTC) criterion to improve the utilization of FTAs. CTC criterion is widely used due to high preference of companies and merits of its own, but sometimes, it causes unintended results because of item classification method of the Harmonized System(HS) table. In order to avoid this problem, a set of products were divided into two groups: a group that can easily meet CTC criterion and; the other with relative difficulties to satisfy the required criterion. This study has proven that for industrial products like machinery and electronic devices, the restrictiveness of CTC criterion is dependent on the parts classification on the HS table. This paper implies that it is useful to assess the impact of parts classification on the restrictiveness of CTC criterion in order to make it sure to set appropriate product-specific rules of origin in FTAs, and thereby to ensure the impact of FTAs to be applied across all industries. Moreover, this paper may be a reference in upgrading the rules of origin in existing FTAs as well as negotiating new FTAs in future.

The global market size of AI based SaMD for medical image in 2023 will be anticipated to reach around 620 billion won (518 million dollars). In order for Korean manufacturers to efficiently obtain CE marking for marketing in the EU countries, the paper is to introduce the recommendation and suggestion of how to reclassify SaMD based on classification rules of MDR because, after introducing the Regulation EU 2017/745, classification rules are quite modified and newly added compared to the Directive 93/42/EEC. In addition, the paper is to provide several rules of MDR that may be applicable to decide the classification of SaMD. Lastly, the paper is to examine and demonstrate various secondary data supported by qualitative data because the paper focuses on the suggestion and recommendation with a public trust on the basis of various secondary data conducted by the analysis of field data. In conclusion, the paper found that the previous classification of SaMD followed by the rule of MDD should be reclassified based on the Regulation EU 2017/745. Therefore, the suggestion and recommendation are useful for Korean manufacturers to comprehend the classification of SaMD for marketing in the EU countries.

When we conduct a data mining procedure on sample data sources, several rules are generated. But some rules are redundant or logically disjoint and therefore they can be removed. We suggest a new rule minimization algorithm inspired from logic synthesis to improve comprehensibility and eliminate redundant rules. The method can merge several relevant rules into one based on data mining and logic synthesis without high loss of accuracy. In case of two or more rules are candidates to be merged, we merge the rules with the attribute having the lowest information gain. To show the proposed method could be a reasonable solution, we applied the proposed approach to a problem domain constructing user preferred ontology in anti-spam systems.

In this study, the unified requirements of IACS on longitudinal strength of ships are verified through the nonlinear time domain analyses in irregular waves. The formula for the horizontal shear force, bending moment and torsional moment for ships of large deck openings are proposed based on the calculation results for existing ships. Also, external hydrodynamic sea pressure, accelerations and motions are calculated using linear strip theory and the corresponding design formulae are proposed. The calculated results are compared with the existing classification rules and will be incorporated into the rules in the near future after more detailed verification.

This paper proposes a GA and Gradient Descent Method-based method for choosing an appropriate set of fuzzy rules for classification problems. The aim of the proposed method is to fond a minimum set of fuzzy rules that can correctly classify all training patterns. The number of inference rules and the shapes of the membership functions in the antecedent part of the fuzzy rules are determined by the genetic algorithms. The real numbers in the consequent parts of the fuzzy rules are obtained through the use of the descent method. A fitness function is used to maximize the number of correctly classified patterns, and to minimize the number of fuzzy rules. A solution obtained by the genetic algorithm is a set of fuzzy rules, and its fitness is determined by the two objectives, in a combinatorial optimization problem. In order to demonstrate the effectiveness of the proposed method, computer simulation results are shown.

In many decision making systems, rule-based approaches are used to solve complex problems in the areas of pattern analysis and computer vision. In this paper, we present methods for generating fuzzy IF-THEN rules automatically from training data for pattern classification and high-level computer vision. The rules are generated by construction minimal approximate fuzzy aggregation networks and then training the networks using gradient descent methods. The training data that represent features are treated as linguistic variables that appear in the antecedent clauses of the rules. Methods to generate the corresponding linguistic labels(values) and their membership functions are presented. In addition, an inference procedure is employed to deduce conclusions from information presented to our rule-base. Two experimental results involving synthetic and real are given.

In this paper, a framework for implementing UFIS (Unified Fuzzy rule-based knowledge Inference System) is presented. First, fuzzy clustering and fuzzy rules deal with the presence of the knowledge in DB (DataBase) and its value is presented with a value between 0 and 1. Second, RDB (Relational DB) and SQL queries provide more flexible functionality fur knowledge management than the conventional non-fuzzy knowledge management systems. Therefore, the obtained fuzzy rules offer the user additional information to be added to the query with the purpose of guiding the search and improving the retrieval in knowledge base and/ or rule base. The framework can be used as DM (Data Mining) and ES (Expert Systems) development and easily integrated with conventional KMS (Knowledge Management Systems) and ES.

The Internet is a packet switched network which offers best-effort service, but current IP network provide enhanced services such Quality of Services, Virtual Private Network (VPN) services, Distribute Firewall and IP Security Gateways. All such services need packet classification for determining the flow. The problem is performing scalable packet classification at wire speeds even as rule databases increase in size. Therefore, this research offer packet classification algorithm that increase classifier performance when working with enlarge rules database by rearrange rule structure into Bitmap Intersection Lookup (BIL) tables. It will use packet's header field for looking up BIL tables and take the result with intersection operation by logical AND. This approach will use simple algorithm and rule structure, it make classifier have high search speed and fast updates.

정확한 객체 분류를 위한 분리 기준을 지정하기 위한 방법은 의사결정수들(decisions trees) 사이에 주요한 이슈이다. 이 논문은 최적의 분리 지점을 찾기 위해 분류에 대한 새로운 분리방법을 기술하고 있다. 모든 시작점 값들(threshold values)을 검색함으로 제공되는 기존의 분리방법들과 다르게, 이 논문에서는 미리 지정된 인터벌들의 확률을 기반으로 하는 분리방법을 제안하고 있다 제시한 방법은 사용자가 인터벌의 수를 조정함으로써 트리의 정확도를 통제할 수 있도록 할 수 있으며, 제기된 분리방법을 이미지 객체 인식하기 위한 특징추출을 수치화함으로써 검색되는 일련의 이미지 데이터에 적용하였다.

The way to assign a splitting criterion for correct object classification is the main issue in all decisions trees. This paper describes new splitting rules for classification in order to find an optimal split point. Unlike the current splitting rules that are provided by searching all threshold values, this paper proposes the splitting rules that we based on the probabilities of pre assigned intervals. Our methodology provides that user can control the accuracy of tree by adjusting the number of intervals. In addition, we applied the proposed splitting rules to a set of image data that was retrieved by parameterized feature extraction to recognize image objects.

With the arrival of the big data era, customer data and data mining analysis have gradually dominated the process of Customer Relationship Management (CRM). This phenomenon indicates that customer data along with the use of information techniques (IT) have become the basis for building a successful CRM strategy. However, some companies can not discover valuable information through a large amount of customer data, which leads to the failure of making appropriate business strategy. Without suitable strategies, the companies may lose the competitive advantage or probably go bankrupt. The purpose of this study is to propose CRM strategies by segmenting customers into VIPs and Non-VIPs and identifying purchase patterns using the the VIPs' transaction data and data mining techniques (K-means clustering and association rules) of online shopping mall in Korea. The results of this paper indicate that 227 customers were segmented into VIPs among 1866 customers. And according to 51,080 transactions data of VIPs, home product and women wear are frequently associated with food, which means that the purchase of home product or women wears mainly affect the purchase of food. Therefore, marketing managers of shopping mall should consider these shopping patterns when they build CRM strategy

The paper presents an automated method for generating fuzzy rules and fuzzy membership functions for pattern classification from training sets of examples and an application to the land-cover classification. Initially, fuzzy subspaces are created from the partitions formed by the minimum and maximum of individual feature values of each class. The initial membership functions are determined according to the generated fuzzy partitions. The fuzzy subspaces are further iteratively partitioned if the user-specified classification performance has not been archived on the training set. Our classifier was trained and tested on patterns consisting of the DN of each band, (XS1, XS2, XS3), extracted from KITSAT-3 multispectral scene. The result represents that our classification method has higher generalization power.

앙상블은 다수의 분류기를 효과적으로 결합하여 분류의 성능을 향상시키는 대표적인 기술이다. 효과적인 앙상블을 위해서는 다양한 특성을 지닌 분류기를 확보하여야 한다. 기존의 앙상블은 개별 분류기의 결과를 바탕으로 분류기 사이의 의존성이나 유사성을 평가하여 분류기 결합을 시도하였다. 따라서 분류기 사이의 유사도의 정확한 측정에 한계를 지니고 있다. 본 연구에서는 이를 극복하기 위해서 다수의 산술연산자 기반 분류규칙을 유전자 프로그래밍을 이용하여 획득하고, 실제 표현형의 유사성을 측정한 후 이를 바탕으로 분류기를 결합한다. 생물정보학에서 많이 사용되는 유전자 데이터 중 하나인 림포마 암 데이터에 제안하는 방법을 적용하여 97% 수준의 높은 분류 성능과 해석 가능한 분류규칙을 획득하였다.

데이터 특성이 연속적이고 애매할 때 퍼지규칙으로 분류 규칙을 표현하는 것은 매우 유용하고 효과적이다. 그러나 일반적으로 정확하지 않은 데이터 특성에 대해서 소속함수를 결정한다는 것은 어려운 일이다. 본 논문에서는 진화알고리즘을 이용하여 효과적인 퍼지 분류 규칙을 자동으로 생성하는 방법을 제안한다. 제안한 방법에서 규칙의 정확성과 이해성을 고려하여 최적화된 소속함수를 생성하기 위해 진화알고리즘을 사용한다. 먼저 지도 군집화로 진화를 위한 초기 소속함수를 생성한다. 진화알고리즘은 전역적 최적 해를 찾는데 효과적이다. 그러나 시간에 대한 효율성이 낮다. 특히 모델 최적화 문제에서는 개체 평가 단계에서 많은 시간이 소요된다. 따라서 본 논문에서는 전체 데이터를 여러 개의 부분 데이터들로 나누고 개체들은 전체 데이터 대신 매번 부분 데이터를 임의적으로 선택하여 개체를 평가함으로써 수행 시간을 단축시킬 수 있는 진화 방법을 제안한다. 제안한 퍼지 분류 규칙 생성 방법의 타당성을 검증하기 위한 실험 데이터로 UCI에서 제공하는 데이터들을 사용하였으며, 실험 결과는 기존 방법에 비해 평균적으로 더 효과적임을 확인하였다.

본 논문은 가중 퍼지소속함수 기반 신경망(Neural Network with Weighted fuzzy Membership Funcstions, NEWFM)을 이용하여 심전도 신호로부터 조기심실수축(Premature Ventricular Contraction, PVC)을 판별하는 퍼지규칙을 추출하고 있다. NEWFM은 자기적응적(self adaptive) 가중 퍼지소속함수를 가지고 주어진 입력 데이터로부터 학습하여 퍼지규칙을 생성하고 이를 기반으로 정상 파형과 PVC 파형을 구분한다. 분류 성능 평가를 위하여 MIT/BIH 부정맥 데이터 베이스를 사용하였으며, NEWFM의 입력은 심전도의 파형에 웨이블릿 변환을 적용하여 추출된 웨이블릿 계수를 사용하였다. 여기에 비중복면적 분산 측정법을 적용하여 중요도가 낮은 계수를 제거하면서 최소의 m 개 특징입력만을 사용한 하이퍼박스로 단순화 시킨다. 이러한 방법으로 추출된 2개의 웨이블릿 계수를 사용한 퍼지규칙은 $96\%$의 PVC 분류성능을 보여준다.

패턴 분류 문제에서 수치적 속성일 경우 퍼지 적용은 효과적인 결과를 보인다는 것은 많은 연구를 통해 알려졌다. 하지만 퍼지를 적용한 패턴분류의 결과는 소속함수의 모양과 개수에 따라 크게 영향을 받는다는 문제점을 가지고 있다. 따라서 이러한 문제점은 퍼지를 쉽게 응용분야에 적용시키지 못하는 원인이 된다. 따라서 본 논문에서는 자동으로 소속함수를 정의할 수 있는 소속함수 학습 방법을 제안한다. 제안한 방법1)은 Penalty연산과 Reward연산을 통해 소속함수가 학습되고 Coverage연산을 통해 소속함수 개수가 학습된다. 제안된 방법의 가능성을 확인하기 위해 벤치마크 데이터 중 Iris, Appendicitis, Breast Cancer를 사용하여 기존 방법과 비교한다.

점차 네트워크상의 침입 시도가 증가되고 다변화되어 침입탐지에 많은 어려움을 주고 있다. 시스템에 새로운 침입에 대한 탐지능력과 다량의 감사데이터의 효율적인 분석을 위해 데이터마이닝 기법이 적용된다. 침입탐지 방법 중 비정상행위 탐지는 모델링된 정상행위에서 벗어나는 행위들을 공격행위로 간주하는 기법이다. 비정상행위 탐지에서 정상행위 모델링을 하기 위해 연관규칙이나 빈발에피소드가 적용되었다. 그러나 이러한 기법들에서는 시간요소를 배제하거나 패턴들의 발생순서만을 다루기 때문에 정확하고 유용한 정보를 제공할 수 없다. 따라서 이 논문에서는 이 문제를 해결할 수 있는 시간연관규칙과 분류규칙을 이용한 비정상행위 탐지 모델을 제안하였다. 즉, 발생되는 패턴의 주기성과 달력표현을 이용, 유용한 시간지식표현을 갖는 시간연관규칙을 이용해 정상행위 프로파일을 생성하였고 이 프로파일에 의해 비정상행위로 간주되는 규칙들을 발견하고 보다 정확한 비정상행위 판별 여부를 결정하기 위해서 분류기법을 적용하였다.