Search

As more and more structured or semistructured data is stored and exchanged in XML format, XML mining becomes increasingly important, especially the study of classification of XML documents becomes more widely. Considering the disadvantage of the current classification of XML documents that based on structure and content, this paper presents an improved method called NM-Similarity computing similarity measure, which maintains an high accuracy rate when XML documents are similar in structure but different in content. This method is applied in KNN (K-Nearest Neighbor) method for classification. The structure similarity between two XML documents is computed by using Euclidean distance, and the content similarity is computed by using Cosine measure. A better result can be seen when classifying XML documents which focus on content more (that is, XML documents are created from the same DTD and the structure is similar) and it is more effective on classifying XML documents. The experiments prove that when XML documents are similar in structure but different in content, NM-Similarity in this paper provides a significant improvement in improving classification accuracy rate. [ABSTRACT FROM AUTHOR]

This paper introduces the Electrocardiogram (ECG) pattern recognition method based on wavelet transform and standard BP neural network classifier. Experiment analyzes wavelet transform of ECG to extract the maximum wavelet coefficients of multi-scale firstly. We then input them to BP classify for different kinds ECG. The experimental result shows that the standard BP neural network classifier's overall pattern recognition rate is well. The ECG in this paper was from MIT-BIH. Experimental result shows that feature vector extracted by the wavelet transform can characterize ECG patterns, and BP neural network classifier has a stronger ECG recognition effect. [ABSTRACT FROM AUTHOR]

Chinese Wall Security Policy (CWSP) is a widely applied access control policy in many fields, especially in commercial world. Delegation is one of the hot topics of access control technologies. Delegation with CWSP means delegation must satisfy not only delegation constrains but CWSP as well. There exist many delegation models, such as RBDM, RDM2000 and PBDM et al, but few focus on it. This paper proposed an approach of how to delegate permission with the restriction of CWSP. Although CWSP is part of delegation constraint, it does not mean that existing delegation models can be easy applied to this kind of delegation. In our approach, we first define two types of delegation constraints consisting of CWSP. Then we discussed different types of revocation and found that automatic revocation can make delegation safer than user revocation. Also, we found that there exists security vulnerability in multi-step delegation and gave some feasible solutions. Finally, this paper gave system implementation architecture and some examples to show how our approach works properly in a situation with CWSP. [ABSTRACT FROM AUTHOR]

This paper introduces a new topic in CSCL field. The paper first analyzes the research questions in interaction and points out that for supporting collaborative learning, the computer does not simply provides a information exchange space for collaborative learning, and more importantly, it should provide the strategic guide and promotion for advancing of collaborative learning acts. Then the paper discusses the process of establishing facilitative strategies in role based discussion in distance CSCL from the perspective of instructional organization strategies. It also explores how to develop CSCL prototype tool based on the strategies proposed in the paper. [ABSTRACT FROM AUTHOR]

Based on rough set, this paper presents a method of Discretization can be used in incomplete continuous information processing. According to the importance of attribute, this method uses the weighted clustering method to determine breakpoints, and chooses breakpoints based on the importances of breakpoints, using the decision-making table 's compatibility as standards. Experiments show that it is a satisfactory alternative method of Discretization . [ABSTRACT FROM AUTHOR]

Pre-classification can effectively improve the performance of handwritten Chinese character recognition. This paper presents a method that uses four corner feature for pre-classification of handwritten Chinese characters. Considering writing variations, we define a set of basic stroke structures and match them with the structures in four corner regions of character image. The matching result will be four feature codes that can be used for character pre-classification. The experiment on 500 Chinese characters from GB2312 shows that our approach can achieve a satisfied result in pre-classification for handwritten Chinese characters. [ABSTRACT FROM AUTHOR]

It is an important issue to analyze the behavioral equivalence of semantic components when studying the dynamic replacement and recombination of them for middleware adaptation. However it is difficult to check the equivalence of behaviors rapidly and precisely. In order to improve the precision of judging, and guarantee the normality and stability of system after replacing and recombining components, the paper mainly uses and extends the theories of equivalence analysis based on higherorder typed p calculus. By analyzing the equivalence of semantic components, it can ensure the behavioral consistency of the new component and replaced component, also effectively guarantee the stability and normality of the whole middleware system's dynamic adaptation. [ABSTRACT FROM AUTHOR]

In order to improve the precision of rating prediction for personalized recommendation online, an approach incorporating personalized contextual information in item-based collaborative filtering is proposed. In this paper we analyze how to learn personalized contextual information and predict ratings for unknown items based on the well-known SlopeOne itembased collaborative filtering. Finally, we experimentally evaluate our results and compare them to the basic Slope One approach. Our experiments suggest that our algorithm provide better quality than Slope One algorithm. [ABSTRACT FROM AUTHOR]

The approximation operator play a vital role in rough set theory. About the approximation operator, usually, we discuss three fundamental elements, including the binary relation in the universe, the basis algebra and the property of approximation operators. In the constructive approach of approximation operators, the properties of the approximation operators depend on the basis algebra and the binary relation. In this paper, we introduce the basis algebra construction approach, named as the basis algebra choosing approach, the basis algebra is built with properties of approximation operators and other certain binary relations. [ABSTRACT FROM AUTHOR]

In the real world, cryptographic computations are often performed on a relatively insecure device which can not be trusted to maintain secrecy of the secret key. Exposure of secret keys is perhaps the most debilitating attack on a cryptosystem since it implies that all security guarantees are lost. The notion of forward security guarantees that the compromise of current secret keys does not compromise past secret keys and past communications. Therefore, it can minimize the resulting damage caused by the exposure of secret keys, and provides a promising approach to deal with the secret key exposure. In this paper, we present an efficient forward-secure public-key encryption scheme and prove it to be secure in the standard model. Most of the performance parameters of our proposed scheme are independent on the total number of time periods. When compared with the previous scheme, our scheme is much more efficient. [ABSTRACT FROM AUTHOR]

The genetic algorithm has very strong overall situation optimization ability, but gradient descent method's s partial optimization ability is stronger. This article has improved the genetic algorithm operation operator, and has figured out a mix optimization algorithm of neural network with their respective merits, which can simultaneously optimize the neural network's structure, also can improve the weight and the threshold value. The results of simulation experiment indicate that this mix optimization algorithm has good approaching ability and pan-ability. [ABSTRACT FROM AUTHOR]

In this paper, we present a localization protocol for wireless sensor networks. The protocol provides three types of localization algorithms, namely, The localization algorithm during the phase of initialization The localization algorithm for new wireless sensor node, and the localization algorithm for moved beacon node. In the first localization algorithm, the trilateration and maximum likelihood estimation are used. During the phase of initialization, the nodes, that their positions had been confirmed, are allowed to send the data packets contained the localization message. The nodes, that their positions are unknown, are not allowed to send any data packet during the phase of initialization, are allowed to receive the data packets from other nodes. During the normal runtime of wireless sensor networks, the new nodes are allowed to send the inquiring data packet. In this localization protocol, the second sign, that is need in TOA(time of arrival), TDOA(time difference of arrival), and RSSI(received signal strength indicator), is not needed. [ABSTRACT FROM AUTHOR]

Nowadays, mostly security solutions are mainly focused on how to defend against various threats, including insider threats and outsider threats, instead of trying to solve security issues from their sources. In this paper, a security requirements engineering process is presented for security requirements elicitation and analysis. And the proposed process will deal with the security requirements at the earlier stages of system design in a systematic way. Thus critical security risks to the business will be considered in the development of the security requirements. [ABSTRACT FROM AUTHOR]

In the process of the restoration of the sub-sampled images, the interpolation template plays a critical role in the performance of the interpolation results. The all-phase filtering templates baesd on the Fourier orthogonal transform superior to the conventional filtering templates. On this basis, this paper introduces an all-phase interpolation method for the sub-sampled images, focusing on the edge region of the images. Our experimental results show the proposed method have higher SNR than the former all-phase filtering method. [ABSTRACT FROM AUTHOR]

Although several efforts were made to support the standardization of Web Services, they are not always perfectly compatible to each other due to the distributed nature. In this paper, a novel approach for process mediation of web services is proposed to adjust the partially compatible messages interchange behaviors to suit the requested/expected interface of each party. Based on the identification of message exchanging sequences in service interactions, several basic process mismatch patterns are presented to develop basic mediator patterns, which can be used to modularly construct advanced mediators that can resolve all possible process mismatches. [ABSTRACT FROM AUTHOR]

Software defect is the main factor influencing software reliability. So detect the software defect in software testing is the most effective way to improve software reliability. In this thesis, we will trying to use MCLP method to predict software defect and do some comparison with some other typical classification algorithm. And the results of our experiments indicate that MCLP is a competitive classification method in the software defect prediction area. [ABSTRACT FROM AUTHOR]

Fuzzy theory is integrated with Artificial Neural Network to create a bridge safety assessment model, through which the Fuzzy-Neural Network is improved in the light of sample data simulation. First, determine network layers interms of the seven critiria for bridge safety assessment. Then enter sample data at the input layer; study sample at the fuzzy reasoning layer by BP calculation method; obtain professional experience and ways of thinking about bridage safety assessment via the network. Finally, compare the assessment results from the network with those from professionals. The comparison proves the artificial fuzzy-neural network's feasibility and efficiency in assessing bridge safety. [ABSTRACT FROM AUTHOR]

Because of the random and complexity of off-line handwritten Chinese characters, the recognition of off-line handwritten Chinese character is always a difficult problem of pattern recognition. The external contour feature which only reflects the outermost feature of the character's strokes has some limitations. Through analyzing the shape of Chinese character, we can extract the contour feature in a specific inner part got by the elastic mesh method, so that the internal and external feature of Chinese character can be mixed into double contour feature. The density feature of strokes of Chinese character will be combined with the double contour feature to form Chinese character recognition feature. At last, support vector machine will be used in training, classification and recognition to show that the method is effective. [ABSTRACT FROM AUTHOR]

In access control, it is a reasonable requirement that authorization mechanism can be implemented intelligently by logic programs. We propose an authorization framework based on logic programs, ranging from design, analysis, and implementation. Our proposed framework is powerful and useful, with RBAC features, flexible authorization, logic-based formalization and integration of policies. It can specify complex security requirements in real world, and facilitate many advantage aspects held by the some prevailing authorization theories. [ABSTRACT FROM AUTHOR]

Yang et al. proposed a cryptanalysis of simple authenticated key agreement protocols that they broke the SAKA-related scheme in 2004. We improved a new scheme which it based on the Tseng's scheme, and it would withstand previous attack such as the modification attack and the dictionary attack. [ABSTRACT FROM AUTHOR]