Knowledge Engineering and Management

In these days, the ever-increasing volume of data requires that data mining algorithms should not only have high accuracy but also have high performance, which is really a challenge for the existing data analysis methods. Traditional algorithms, such as classification and feature selection under neighborhood rough sets, have been proved to be very effective in real applications. Parallel approach to these traditional algorithms could be a way to take the challenge. This is what we present in this paper, the design of parallel approaches to neighborhood rough sets and the implementation of classification and feature selection. Two optimizing strategies are proposed to improve the performance of the approaches: (1) The distributed cache is used to reduce I/O time. (2) Most of computations are put into the Map phase which helps reduce the overhead of communication. The experimental results show that the proposed algorithms scale pretty well and the speedup is getting higher with the increasing size of data.

Based on the text subjective judgment algorithm based on the rough set, we proposed an improved logarithmic linear model and fuzzy set combining the subjective intensity of learning method Chinese words and lexical subjectivity recognition, which is applied in the E-learning interactive text, and achieved better recognition results.

Geometry theorem proofs like propositions in Euclid’s geometry elements contain fruitful geometry knowledge, and the statements of geometry proofs are almost structural mathematics language. Hence, it is possible to let computer understand geometry theorem proofs. Based on the process ontology, a novel geometry knowledge base (GKB) in this paper is built by letting computer learn from theorem proofs. The resulting process ontology is automatically constructed by extracting abstract and instance models (IMS) from proofs. The abstract model displays the causal relations of conditions with conclusions, and the instance model (IM) holds the formal relationship of abstract model so that the deduction can be reused. Thus, two kinds of models completely describe the proving process of geometry theorem. Furthermore, GKB based on the process ontology can be gradually extended by learning from more and more proofs. Finally, GKB learning from about 200 examples is implemented, and an application in automated theorem proving is given.

This paper studies return of subscribing for new shares in Chinese Growth Enterprise Market (GEM) based on random simulation, and draws some useful conclusions. One is that investors can make more profit by selling their new shares at closing price than at opening price in the first trading day, the other is that percentage increase of a new stock in the first trading day has a bigger impact than the rate of successful subscription for new shares on return of subscribing for new shares in the Chinese GEM. The paper also finds both offering price to earnings ratio and amount of capital raised by initial public offerings (IPO) have significant negative effect on return of subscription for new shares by building a linear regression model. The paper is useful for investors to forecast return and control risk when they subscribe for new shares in Chinese GEM.

The rough set model over dual-universes (RSMDU) is a generalized model of the classical rough sets theory (RST) on the two universes. It is an effective way to use incremental updating approximations method in the dynamic environment to better support data mining-related tasks based on RST. In this paper, we propose an incremental learning approach for updating approximations in RSMDU when the objects of two universes vary with time. An illustration is employed to show the validation of the presented method.

As it is difficult to describe the mechanism of ship repair risk, common methods are less credible for its assessment. By analyzing ship repair risk, this paper identifies the causes and consequences of risk, and utilizes frame-based representation to construct the case representation for ship repair risk. In addition, similarity functions are developed for enumerated attributes, numeric attributes, and fuzzy attributes, so as to perform reasoning of repair risk from the approach of

K

-Nearest Neighbor. Case analysis presents that the application of case-based reasoning in ship repair risk assessment is easy to understand and extend.

The Dempster–Shafer theory of evidence for attribute aggregation provides a method to deal with uncertainty reasoning. In this paper, uncertainty reasoning method based on rule-base with certainty interval is investigated. First, knowledge representation with interval uncertainty is defined and the matching principle is given. Then, a rule-based inference method under interval numbers using Dempster–Shafer theory is derived. A numerical example is examined to show the implementation process of the proposed method.

How to predict the topic diffusion is a challenging research work in social media data mining. The classical research works in Twitter and Micorblog mainly focus on diffusion links that ignore the importance of diffusion content. In this paper, we propose a Link Information Flow-based topic diffusion prediction model, which combines the link view and content view in diffusion. The experiment results show that our model achieves good performance in topic diffusion prediction.

In the purpose to solve the problem of difficult construction, poor effect, complex update, and poorly comprehensive presentation, the paper proposes a method for the knowledge map construct model based on knowledge requirements and knowledge connection. According to the task context, product structure, and the personal knowledge structure, this method can get the knowledge requirements of the users to generate the knowledge retrieval expressions to obtain the knowledge points. Then it can construct the multi-dimensional dynamic local knowledge map through the analysis of multiple dimensions and using the design of aircraft landing gear as an example to verify the feasibility of this method.

In Traditional Chinese Medicine (TCM), the prescription is the crystallization of clinical experience of doctors, which is the main way to cure diseases in China for thousands of years. The relationship between prescriptions is the important research field of the pharmacology of traditional Chinese medical formulae. In this paper, we present a system which mines the relationship between TCM prescriptions from prescription literatures, including the composition and function of prescriptions. The relationship of prescription composition is established by Trie-based tree, and the relationship of prescription function is established by the topic model. The evolution of prescriptions is presented to users in a visualization way. Finally, the experiment validates the effectiveness of our method.

In this paper, a novel

hiding

-

missing

-

artificial utility

(HMAU) algorithm is proposed to hide sensitive itemsets through transaction deletion. It considers three side effects of hiding failure, missing itemsets, and artificial itemsets for evaluating whether the processed transactions are required to be deleted or not, in sanitization process. Experiments show that the proposed HMAU algorithm has better performance whether in the execution times, the number of deleted transactions, and the number of side effects.

Concept relatedness is widely used in information retrieval, text classification, semantic extension, and other fields. So measuring the concept relatedness efficiently is an important task. Previous studies rarely distinguish between relatedness and similarity; they usually use a common formula. We suggest that concept relatedness consists of similarity and relevance, which should be computed differently. In this paper, we first give a similarity measure based on path length, taxonomy depth, and different relations between concepts. Then we propose a method to measure the specific association relation besides basic relations. Finally, incorporating both similarity and specific relevance, we get an overall formula of computing concept relatedness. Compared to existing methods, our measure of concept relatedness is more consistent with human judgment.

As academic relationship networks are implied in various academic activities which relate to study and work experiences of scholars, extracting the transaction information of academic activities is important for mining academic relations networks. For the characteristics of long-distance dependencies in Chinese sentences of academic activities, this paper proposes to extract the transaction information from Chinese documents by using sequence forecast based on Conditional Random Fields (CRF). Often, in research project applications, the resumes of applicant and team members include many complex Chinese sentences about academic activities. We design novel methods to analyze special sentence patterns in those resumes. More specifically, we focus on the design of feature templates according to the sentences characteristics of academic activities, and employ the regular matching method to deal with inaccurate words segmentation, especially for academic-specific words. Through evaluating tests, we choose the optimum feature templates and input to CRF++ model to label trunk words of the sentences. The transaction information extraction of academic activities is implemented. Experimental results show the effectiveness of the proposed approach.

With the rapid development of Internet and Web2.0, product review as a kind of user generated content plays an important role in people’s lives. However, an overwhelming amount of reviews exist and they vary significantly in quality. The purpose of this paper is to alleviate this problem of information overload. We adopt a regression approach to predict the helpfulness of book reviews. A high quality subset of reviews is then selected accordingly. Since book review is typically longer and more diverse in content and style, we extract some complicate text features and propose a new way to weight word features. Experiment shows that our method performs better than the baseline TFIDF model.

Word Sense Disambiguation (WSD) has been an important and difficult problem in Natural Language Processing (NLP) for years. This paper proposes a novel WSD method which expands the knowledge for senses of ambiguous word through semantic knowledge in WordNet. First, selecting feature words through syntactic parsing. Second, expanding the knowledge for the ambiguous word senses through glosses and structured semantic relations in WordNet. Third, computing the semantic relevancy between ambiguous word and context and achieving the purpose of WSD by semantic network in WordNet. Lastly, adopting the Senseval-3 all words data sets as the test set to evaluate our approach. Through a detailed experimental evaluation, the result shows that our approach achieves improvements over some classical methods.

Aiming at the problem of the low utilization rate of knowledge, low rate of staff dependence on knowledge management system, and the problem of lack of effective evaluation system and method, a comprehensive evaluation method of enterprise knowledge and employees was proposed. The knowledge evaluation system of multilevel and index was established based on the characteristics of knowledge resources and employee behaviors in knowledge management system. And then a knowledge evaluation model was constructed. The knowledge utility was used to represent the total index of knowledge. The method divided knowledge utility into knowledge quality and knowledge influence. The individual total score was divided into participation and contribution used to represent the total index of staff. According to this model, evaluation algorithm of each index was put forward. Finally, an example test was used to demonstrate the effectiveness and feasibility of the proposed method.

Aiming at the problem of the low utilization rate of knowledge in product lifecycle, and the problem of the knowledge that cannot be take full use to serve the business process, a construction scheme of the knowledge service oriented to product lifecycle was proposed. Firstly, the demands of knowledge service oriented to product lifecycle were analyzed and the basic concepts and characteristics of knowledge service were put forward. Secondly, the model of knowledge service and its structure was built; the integration of the system, the acquisition of the multi-source heterogeneous knowledge, and the technologies of knowledge push have been analyzed. Finally, part of the system function was given to illustrate the feasibility of the proposed method.

There is an intrinsic correlation between technological diversification and enterprise product strategy. In this study, on the basis of product strategy, the impact on innovation performance of technology diversification is investigated and practical data of product innovation is also analyzed by structural equation modeling (SEM) methods. The research found that the intermediary role of product strategy is remarkable in technology diversification impacts on both business performance and innovation performance, especially significant on business performance.

This paper proposes a method based on concept expansion to extract paraphrase collocation. Collocations of forms of

$$ \left\langle {{\text{V,}}\;{\text{OBJ,}}\;{\text{N}}} \right\rangle $$

V,

OBJ,

N

(verb-object collocations) and

$$ \left\langle {{\text{V,}}\;{\text{SUB,}}\;{\text{N}}} \right\rangle $$

V,

SUB,

N

(subject-predicate collocations) are extracted after syntactic analysis is done to the sentences. Then the words used in the collocations are expanded based on related words getting from concept semantic to get the candidate of paraphrase collocations. In order to filter these paraphrase collocations, following four features are chosen: part of speech feature, mutual information feature, HowNet-based semantic similarity feature, and context-based semantic similarity feature. Compared to existed method, this method does not restrict the word in paraphrase collocation to synonym. The experiment shows that every feature exploited is useful for improving the performance.

Since WEB2.0, more and more online communities began to use tag—words selected by users or generated by computer algorithms—to help people find or organize data resources. Unfortunately, the tags are generally in a random order without any importance or relevance in information, which seriously limit the effectiveness of these tags in tag-based applications. In this paper we present a tag ranking method which first computes the probability of each tag associated with a given book, and then adjust the probability as well as the tags’ order based on users’ tag-click behaviors. Then an initial strategy which provides a better initial probability is described to improve our method. Experimental results show that users’ tag-click behaviors can reflect the relevance between books and tags to some extent and our approach is both efficient and effective.

Considering the insufficient knowledge management and technology innovation in the domestic construction industry, this paper proposes the design framework of the technology innovation platform based on TRIZ by taking advantage of available patent knowledge in the industry. The function modules and their roles in technology innovation are illustrated through the graphic user interfaces and the underlying database. Based on extracted construction patent knowledge, the development of construction technology innovation platform enables a heuristic environment to help the industry improve the innovation capacity and efficiency by motivating knowledge worker’s innovative thinking.

Mutual information (MI) algorithm has many shortages in the feature selection of text categorization compared to other selection algorithms. For these shortages, this article introduces some important factors like term frequency or something that MI has not yet considered, and then puts forward the improved MI algorithm based on the composite ratio factor. And by the experiment the improved method can get a good improvement effect.

Cloud computing is an emerging new computing paradigm for delivering some computing services to consumers, which provides scalable and inexpensive service-oriented computing infrastructures with good quality of service levels. In this paper, based on the studies about the characteristic of knowledge service, a feature model of knowledge service is proposed with three features, such as service requirement, knowledge service process, and the quality of knowledge service (QoKS). Furthermore, the architecture about service-oriented knowledge acquisition is built to provide a fundamental for knowledge service, and the lifecycle process about knowledge service is studied under the service-oriented architecture. Then, a best practice about Knowledge Cloud, named Eknoware, is developed based on the architecture above-mentioned, which can provide some knowledge service patterns and reorganize the knowledge clusters to suitable customers.

A detailed visual analysis of publications related to crop growth model development and trend was carried out based on the latest version of information visualization and analysis software CiteSpace II. A total of 6,079 publication data between 1995 and 2011 were collected from Thomson ISI’s SCI (Web of science in the Science Citation Index Expanded Edition). After analysis by CiteSpace, the most productive countries related to Crop Growth Model, as well as institutes, key scholars, co-citation patterns, etc., were visualized and identified from 1995 to 2011.

Providing quality aware services to customers is an important problem in the composition of Web services. Often, the quality of services is considered from different aspects that we call criterions. Based on the criterions, customers compare and choose better (composed) Web services. Previous methods compare Web services by comparing their aggregate values, which are weighted sum of each quality (the

WSum model

). However, the WSum model is not proper to support qualitative preferences over criterions. Herein, we proposed a new evaluation model that lexically compares the qualities of Web services, which is called the

Lexical model

. In contrast with WSum model, our proposed Lexical model is more powerful in modeling users’ preferences both quantitatively and qualitatively. We will prove that it is in

O(k

2

)

for a WSum model to simulate the Lexical model, where

k

is the number of criterions. We may note that though the idea of Lexical models has been utilized in other areas, few researchers in the field of Web service composition noticed the way of using the Lexical model introduced here.

This paper introduces a content management system (CMS) for enterprises, which combines the enterprise resources with the information-issuing website, and which integrates content collection, content creation, content editing, template production, content approval, content issuing, and content browsing applications into a whole. This system can make users easily and efficiently issue, manage, and exchange information via websites according to users’ different demands, indicating that the efficiency of website daily information processing could be improved. This paper elaborates the architecture of the enterprise content management system, presents its major functions and key technologies, and gives the website applications of this system as the example.

The objective of this research was to assess soluble solids content (SSC) of Cuiguan pears by visible/near infrared (Vis/NIR) and variable selection methods. Vis/NIR transmission spectra of Cuiguan pears were taken by a USB4000 spectrometer. A variety of pretreatment methods and three variable selection methods were used to select important variables in this study. After that, Partial least squares (PLS) was used to develop calibration model. The results show that competitive adaptive reweighted sampling (CARS) is an effective variable selection method for SSC of Cuiguan pears. PLS using the selected variables by CARS combined with multiplicative scattering correction (MSC) obtains the best results. Compared with full spectrum PLS, The number of variables in MSC-CARS-PLS model reduces from 1,400 to 84, the correlation coefficient rises from 0.88 to 0.96, and the root mean square error decreases to 0.29 °Brix.

In order to represent the design knowledge of complex product, logical descriptions of different knowledge are conducted by character analysis for the generalized design knowledge. Then a knowledge representation system is proposed in this chapter utilizing the three-layer structure of “design object ontology—design process ontology—data resource” (OPR). And the semantic description frame is defined based on the suggested upper merged ontology (SUMO) to implement the ontology collaborative construction using the ontology web language (OWL). Based on the constructed ontology, the storage structure for generalized knowledge is determined to actualize the generalized knowledge representation. Finally, the feasibility and effectiveness of the proposed technique in this chapter is demonstrated by the design knowledge representation for the self-propelled gun.

In Group Corporation, the large number of distributed and heterogeneous knowledge resources makes it difficult to give full play to their integrated benefit. Knowledge Fusion is proposed to solve this problem. Aiming at the specific characters of Group Corporation, this paper presents a method for Multiple Domain Ontology Construction. Dynamic ontology is a temporary ontology built on the base of Multiple Domain Ontology to compact the knowledge requirements and form the target knowledge resources. A Dynamic Ontology Construction Method is also proposed by analyzing knowledge requirements for more effective Knowledge Fusion. At the end of the paper, an application example of Dynamic Ontology Construction is presented to prove the feasibility of the method.

As it is known to all, many factors may have influence on the movement of stock index. In stock index forecasting, how many quantitative indicators should be introduced in order to obtain the best forecasting result? And is it true that more indicators translate into higher forecasting accuracy? These issues have long been puzzling to researchers of stock index forecasting. In this paper, we carried out data mining on some quantitative indicators with influence on the movement of stock index, then we had short-term forecasting of Shanghai Component Stock Index with BP+GA model. Results of our research are as follows: forecasting with combination of indicators has better result than forecasting with single indicators; combinations of indicators through selection and optimization have the best result; more indicators introduced into forecasting model do not translate into higher accuracy. The results of our research in this paper demonstrate the necessity and significance of data mining in stock index forecasting.

Design rationale (DR) explains why an artifact is designed the way it is, which is a kind of important trait knowledge. Successful reusing of DR knowledge plays an important role in guiding design thinking and could raise design efficiency and design quality significantly. In order to satisfy the rising requirements of DR reuse, a formal expression of DR is described and a semantics-based design rationale reuse model based on the expression is proposed. Designers’ knowledge needs are analyzed based on semantics, and isolated DR elements retrieved are extended into DR segments by reasoning. After putting any isolated DR nodes in the context of design, the DR segment is easier to understand. Finally, a prototype system is developed and several examples are given to prove the validity of proposed methods.

This paper proposes an algorithm to represent them in a granular way—information granules. Information granules can be regarded as a collection of conceptual landmarks using which people can view the data and describe them in a semantic way. The key objective of this paper is to introduce a new granular way of data analysis through their granulation. Several experiments are done with synthetic data and the results show a clear way how our algorithm performs.

This paper presents the effect of control on knowledge transfer in service supply chain with an empirical data from China Mobile. Formal control and Social control are mechanisms usually taken to promote knowledge transfer by buyer in service supply chain. So, the role of formal control, social control and their complement to knowledge transfer are studied in this paper. Using the data collected from China Mobile, we found formal control is positively related to knowledge transfer, and social control is also positively related to knowledge transfer. In addition, there is a complement effect between formal control and social control to knowledge transfer. Theoretical and managerial implications are discussed.

This paper provides an alternative way of performance measurement on production efficiency using optimization method. This paper provides an alternative way of performance measurement on production efficiency using optimization method. It is believed that any network production can be transformed into a series system where each stage in the series has a parallel structure. Thus, we first present models for simple series structure and parallel structure production under evaluation, separately. Then the models for general network structure production that consists of both the series and parallel structures are given. Performance efficiency including the relationship between the whole production and its divisions in series structure, parallel structure and general network structure are discussed.

This article is based on the emotional cognition and learning style of the ECLS model, giving a reasoning tools-Dynamical Uncertainty Causality Graph (DUCG) that the teaching strategies reasoning needs, then it uses DUCG for network teaching reasoning based on the emotional cognitive interaction.

In many fields, labeling samples is a time-consuming and costly work. This paper describes an automatically labeling samples method based on SVDD with transductive confidence (TC-SVDD). The new algorithm labeling samples automatically by lead transductive confidence idea into support vector data description. It gives the confidence lever about labeling result to improving the labeled samples quality. Experiment results on UCI data sets show the algorithm has advantages on label samples with high quality.

The objective of this research is to improve soluble solids content (SSC) prediction in navel oranges by visible/near infrared (Vis/NIR) semi-transmission spectra and uninformative variable elimination-genetic algorithm-least squares support vector machine (UVE-GA-LSSVM). Spectra of navel oranges were acquired using a QualitySpec spectrometer in the wavelength range of 350 ~ 1,000 nm. After applying spectral pretreatment methods, UVE-GA was used to select variables, then LSSVM with three kernel functions (RBF kernel, linear kernel, polynomial kernel) was used to develop calibration models. The results indicate that Vis/NIR semi-transmission spectra combined with UVE-GA-LSSVM has good performance on assessing SSC of navel oranges, and SSC is improved. The

R

2

s and RMSEPs of SSC for RBF kernel, linear kernel, and polynomial kernel in prediction set are 0.850, 0.848, 0.849 and 0.419, 0.421, 0.420 %, respectively.

In Digital Library, information is often stored in unstructured and semi-structured textual form. Domain modeling techniques are used for specific domain knowledge services in order to make use of the massive amounts of textual information. Ontology is an efficient way to build specific domain model for abstraction of concepts and relations. In this , we propose an ontology-based domain modeling framework, CADAL-ODM, to structure domain model. CADAL-ODM is designed to provide multi-level and multi-granularity knowledge services in order to meet the various requirements of users in digital library instead of basic reading service. We explain our work by which knowledge is extracted automatically from the unstructured and semi-structured documents. Specific domain model is structured by specific domain ontology description and extracted knowledge. We evaluate the framework with real data sets, specifically, the medical records of TCM documents set from CADAL.

This paper established and designed evaluation index system based on “Information Technology application fields—service quality promotion” two dimensions model. The Information Technology application fields included academic and non-academic student support services. The service quality promotion included tangibles, reliability, responsiveness, assurance and empathy. This two dimensions framework evaluation method is helpful to understand the performances on specific aspects of student support services, and then do targeted improvement for the distance educational institutions.

Now there exists a lot of information from Internet and the information implies abundant knowledge. But the information is usually uncertain, imprecise, incomplete, coarse, and vague. So it is very difficult to acquire knowledge from the miscellaneous information. One of the effective methods to process this kind of information is Rough Set theory. As an important method of the acquisition and process for knowledge, Rough Set theory has given the algebraic representation for knowledge by means of the equivalence relations of algebra and the inclusion relations of set theory. But this representation makes knowledge understanding difficult. In order to overcome this problem the concept of granulating is proposed and the granular representation for knowledge is suggested. The granular representation for knowledge makes complicated problems simplified. It is nearer to the thinking habits of mankind and it can describe the roughness of knowledge quantitatively. This chapter discusses the algebraic and granular representations for knowledge, respectively. Some definitions, properties, and theorems under two different representations are analyzed. The research results justified that two different representations for knowledge are equivalent, but the granular representation is more direct and easier to be understood.