由于人类活动的复杂性和多维性,活动模式的挖掘具有很大的挑战性.本文提出了一个基于时序活动序列计算用户之间的相似度,通过聚类分析来挖掘活动模式和社会人口学模式的方法,对模式进行趋势分析.实验结果表明,提出O(p(m–p))的相似度算法,可以有效地进行聚类.在此基础上,通过时序活动图和概率密度函数(PDF)图的可视化以及统计分析,挖掘出了活动及其社会人口学模式,然后通过对连续多年的数据挖掘,获取行为及其社会人口学模式的发展趋势,以此得到了相似的活动行为具有相似的社会人口学特征的结论.

针对分布式光伏电站阴影遮挡提出一种基于随机森林算法的人机协同判别方法.通过遮挡机理分析和逆变器遥测参数转换构建组串直流侧电流离散率、太阳高度角、太阳方位角及电站瞬时发电水平等关键特征参数,搭建随机森林遮挡诊断模型.基于网格搜索法和K折交叉验证法优化参数,通过准确率对比确定基于信息增益的分裂方式.对比支持向量、逻辑回归及决策树等主流算法模型,发现随机森林算法在遮挡诊断场景中具有较强的优势,结合专家系统得出诊断方位后,现场验证了"基于信息增益的随机森林和专家系统"方法的有效性.

Domain Adaptation (DA) is a learning problem where we wish to leverage labeled data from source domain to learn a hypothesis performing well on a different but related target domain for which no labeled data is available. In this paper, an idea of unified method named UNDA (United Domain Adaptation) is used for DA problems. The main contents are: (1) Find a feature mapping function to bridge the gap between two domains, where feature representation is learned from the source domain for the target domain. (2) Find a discriminant function that is adapted from the source domain instances to the target domain instances step to step, where iteratively deleting source domain instances and adding target domain instances. Experimental results on two data sets show that our new approaches can be successfully used to perform domain adaptation for text and sentiment classification problems. © 2013 Binary Information Press.

This paper focuses on the problem of opinion summarization of customer reviews, which has not been solved completely in how to get a summary that overlaps the most important topics of review text and to extract the opinion on these topics. With the intuition that both factual text and evaluative text should be considered in opinion summarization, this paper deals with sentiment and topics of review text simultaneously and proposes a multi-document opinion summarization approach based on Latent Dirichlet Allocation model. Experiments on product reviews show that the approach is effective.

Most of the previous researches on multi-document summarization concentrate on factual text. How to get a summary that cover the most important topics of evaluative text and to extract the main opinion on these topics have not been solved completely yet. This paper considers both sentiment and mixture of topics of a sentence simultaneously and proposes a multi-document opinion summarization approach based on Latent Dirichlet Allocation model and semantic orientation of word. Experiments on product reviews show that this approach performs better.

In order to improve innovation ability,interesting and scientific attitude better,bilingual teaching is helpful for students to read the related foreign literature,familiar with representation of professional knowledge.And bilingual teaching also facilitate the student to express professional knowledge and communicate with foreign researchers in foreign language.In this paper we make an investigation on the goal of bilingual education,usage of teaching materials,education means and methods,education content according to our course construction and experience of bilingual education.

With the continuous growth in the XML data, the ability to search in massive collections of XML data becomes important. In this paper, we present efficient techniques that are able to employ bloom-filtering to decrease computation complexity that is used to filter irrelevant XML paths. After filtering, a kind of semantic measure is used to compute similarity between the query and the relevant XML documents, which is used to rank retrieval results. Experiment results show that the retrieval prototype system based on bloom-filtering runs faster than ever under the almost same average precise.

Transitivity is an important property of relationship used in logical reasoning. The paper proposes a new theory that the transitivity of Part-Whole (PW) relation is determined by semantic constraints. In our idea, PW sub-relation is a kind of PW relation with specific semantics, which adds the semantic constraints on PW relation. By adding new constraint, the PW sub-relations can be subdivision until get taxonomy of PW relation. The semantic constraints are determinant factor of the transitivity of PW relations. The paper lists the main semantic constraints and analyzes how they affect the transitivity of PW relations. © 2011 by Binary Information Press.

Most of the previous researches on sentiment analysis concentrate on the binary distinction of positive vs. negative. This paper presents the multi-class sentiment classification problem that attempt to mine the implied rating information from reviews. We use four machine learning methods and two feature selection methods to find out whether or not the multi-class sentiment classification problem is the same to the binary sentiment classification problem, and whether it is equal to the traditional multi-class classification problem. Experiments show that multi-class sentiment classification problem is difficult than that of only determining the polarity of a review and that it is different from traditional multi-class classification problem, thus traditional multi-class classification method can not be directly used to deal with this problem.

With the widespread diffusion of semi-structured data in XML format, algorithms for mining information from XML documents are becoming increasingly important. So a similarity function is the key of a successful XML data management process. In this paper, we propose a new method to measure the similarity between XML documents by considering their structures and contents, which comprises three layer matching: element matching, path matching and document matching. The similarity of document's structure is found by partial matching technique and that of document's contents is found by taking into account of the syntactic information, semantic information and position of elements.

The categorization of deep web resources is a key operation in organizing deep web resources. Aiming at the problem of lacking semantic support with query probing in the research of Deep Web databases classification, this paper proposes a novel classification approach based on ontology. The main idea is following: Firstly, the characteristics of category are represented by domain ontology, and the instances from domain ontology are used as query terms to probe the Deep Web. Their retrieved data after probing are created a context document, where such a context document contains many characteristics of categories. Next similarity between the context document and categories is computed using Earth Mover's Distance (EMD) method; finally the context document associated with a query is evaluated to identify which category that this context document belongs to, and the context document in response to Deep Web query interface is assigned to categories according to the similarity score. The experiments show that the new approaches achieve better precisions and recalls compared with random probing post-query and pre-query approaches. © 2009 Binary Information Press.

Both of XML document and user's query are represented by the set of paths from the root node to leaf nodes. So the context and content information contained in the corresponding path is a vital important clue to research XML retrieval. This paper presents an approach, NPathSim, for measuring similarity between two paths. XML Path retrieval was performed to evaluate the performance of NPathSim. The experiments show that our path retrieval method can achieve better performance than other methods.

Computing the similarity between two words plays an important role in applications such as word sense disambiguation and automatic correction of word errors. However, along with the development of ontology, how to compute the semantic similarity between two words based on ontology has become a new issue. As WordNet is an excellent general ontology, this paper proposes a fuzzy approach for measuring the semantic similarity between a pair of words in WordNet, with a formal mathematical foundation. We firstly consider sufficiently the structural information of words in WordNet and represent these words by fuzzy sets. Then we define degree of membership of the fuzzy sets. Finally, we compute the fuzzy similarity between two fuzzy sets corresponding to two words. And experimental evaluation demonstrates that the proposed measure outperforms traditional similarity measures on a common testbed. 1548-7741/ Copyright © 2009 Binary Information Press.

Clustering XML documents semantically has become a major challenge in XML data managements. The key research issue is to find the similarity functions of XML documents. However, previous work gave more importance to the topology structure than to the semantic information. In this paper, the computation of similarity between two XML documents is based on both structural and semantic information. Then a minimal spanning tree clustering method is used to cluster XML documents. The experiment results show that the new method performs better than baseline similarity measure in terms of purity and rand index.

The data model of XML document can be labeled as a tag tree of element nodes. Such tree model can be represented by the set of paths from the root node to leaf nodes, which describes the structure of XML document. This paper presents an approach for measuring similarity between two XML paths that consists of (1) ElementSim, a similarity function specifically designed for measuring linguistic similarity between two elements in two different paths, which take into account both semantic and syntactical information of elements. (2) NPathSim, a similarity function specifically designed for measuring similarity between two paths, which combines both the linguistic similarity between elements and the context descriptions of paths. Path retrieval was performed to evaluate the quality of NPathSim. The experiments show the proposed similarity approach can achieve higher quality on XML data set. ©2009 IEEE.

Deep Web database clustering is a key operation in organizing Deep Web resources. Cosine similarity in Vector Space Model (VSM) is used as the similarity computation in traditional ways. However it cannot denote the semantic similarity between the contents of two databases. In this paper how to cluster Deep Web databases semantically is discussed. Firstly, a fuzzy semantic measure, which integrates ontology and fuzzy set theory to compute semantic similarity between the visible features of two Deep Web forms, is proposed, and then a hybrid Particle Swarm Optimization (PSO) algorithm is provided for Deep Web databases clustering. Finally the clustering results are evaluated according to Average Similarity of Document to the Cluster Centroid (ASDC) and Rand Index (RI). Experiments show that: 1) the hybrid PSO approach has the higher ASDC values than those based on PSO and K-Means approaches. It means the hybrid PSO approach has the higher intra cluster similarity and lowest inter cluster similarity; 2) the clustering results based on fuzzy semantic similarity have higher ASDC values and higher RI values than those based on cosine similarity. It reflects the conclusion that the fuzzy semantic similarity approach can explore latent semantics.

Focused crawling is increasingly seen as a solution to address the scalability limitations of existing general-purpose search engines, by traversing the Web to only gather pages that are relevant to a specific topic. How to predict the relevance of the unvisited pages pointed to by candidate URLs in the crawling frontier to a given topic is a key issue in the design of focused crawlers. In this paper, we propose a novel approach based on multiple relevance prediction strategies to address this problem. For cross-language crawling, we first introduce a hierarchical taxonomy to describe topics in both English and Chinese. We then present a formal description of the relevance predicting process and discuss four strategies that make use of page contents, anchor texts, URL addresses and link types of Web pages, respectively, to evaluate the relevance more accurately, in which we propose a particular strategy using Chinese URL addresses to estimate the relevance of cross-language Web pages. Finally, we get a new focused crawling algorithm (FCMRPS, Focused Crawling based on Multiple Relevance Prediction Strategies) based on the combination of these strategies and Shark-Search, which is a classic focused crawling algorithm. Experiments show that the FCMRPS is more effective than the traditional algorithms, namely Breadth-First, Best-First and Shark-Search, in terms of precision and sum of information.

Wikipedia (Wiki) is a collaborative on-line encyclopedia, where web users are able to share their knowledge about a certain topic. How to make use of the rich knowledge in the Wiki is a big challenge. In this paper we propose a method to construct domain ontology from the Chinese Wiki automatically. The main idea in this paper is based on the entry segmenting and Feature Text (FT) extracting, where we segment the name of entries and establish the concept hierarchy firstly. Secondly, we extract the FTs from the descriptions of entries to eliminate the redundant information. Finally we calculate the similarity between pairs of FTs to revise the concept hierarchy and gain non-taxonomy relations between concepts. The primary experiment indicates that our method is useful for Chinese domain ontology construction.

According to the characteristics of typical power unit above 200MW,the thermo-economics diagnosing models were set up.Based on these models,a thermo-economics diagnosing system with good generalization was developed including the functions such as data initializing,thermo-economics diagnosis,optimal operation level diagnosis,thermodynamic test and daily operation management.The example and practice proved that this system was an exact,quick and general tool to improve thermo-economics and energy conservation of power unit,which could enhance the operation management level of thermal power plants.

Sentiment classification is a burgeoning field in web information search community with the explosion in the number of various online forums, weblogs and various web applications, which intends the activity of recognizing, classifying, extracting and annotating the expression of sentiment, opinion and affect in some content. From 1990s, a lot of researches have been carried on and various methods have been proposed. In this paper, the up-to-date methods for sentiment classification are studied and the latest progress of sentiment classification is reviewed. After giving a brief review of sentiment classification, three main approaches of current research, the semantic orientation approach, the machine learning approach and the ontology-based approach, are described and compared. At last, some challenges of sentiment classification are discussed, together with the potential future directions.