Considering the effect of desk and chair on the pedestrian moving velocity,a new cellular automaton evacuation model with seat andaisle regions is proposed,and the evacuation time with the occupant density,different exit positions and internal layout are analyzed.Simulation results show that:when the exit is in the side,it should not be directly opposite to the desks andseats;it is very beneficial when the aisle is close to the exit side of the wall;and the number and width of the aisle will effect on the evacuation time.In order to achieve theoptimal evacuation efficiency,the internal layout should make pedestrians leave the seating region more quickly and avoid the congestion with the formation of density in the exit.

A measure of the velocity fluctuations during a tripis defined as the acceleration noise,and it can be used as a quantitativeevaluation index for riding comfort.Acceleration noise is mainly affected by three factors:driver,road and traffic conditions.In this paper,a new acceleration noise model based on the horizontal、vertical and lateral acceleration is created.The new model not only can be used to analyze the safety of the traffic,but also can provide data for highway design.

Aiming at the problem that tunnel lighting wastes a lot of electric energy,after analyzing the current control model of lighting in tunnel.The control constructure of the system and the design of the fuzzy controller aredescribed.In this system,the traffic flow and illuminance out of the tunnelare input,and the LED is used as lighting source.During the process,the adaptive control is realized,and continuous adjustment of tunnel illuminace is achieved,which shows that this system saves energy.The illuminance curve in the tunnel is comparedthrough simulation with those of different illuminance out of the tunnel and different amount of trofic.The result shows that the method is effective.

Based on characteristics of chaotic signals,this paper presents a secure communication method which contains synchronization of chaotic systems.Itdiscusses the design of multi-level chaotic modulation system which increases the decoding difficulty ofinformation.For receiver,pseudo-signal is not useless because it contains state variables that synchronization needs.The methodis explained and demonstrated in details in the paper.Moreover,numerical simulation shows the effectiveness of the method.The noise that exists in practice isalso discussed to show the robustness of the method of secure communication provided in this paper.

Let f:X→X and f be the set-valued mapwhich is induced by f .In the paper,it is proved that if f is topologically ergodic,it implies f is also topologically ergodic,the converse is not true,and the topological ergodicity of f and f is equilant under W^e-topology,then f is topologically ergodic if and only if f_mis topologically ergodic.It is also proved that,for some integerk,if f^k is chain ergodic,it implies f is also chain ergodic.Let f have POTP,7 equilant condtions on f which is chain ergodic are given.

The aim of this paper is to disscuse some characteristics of the core inverse of matrix multiplication corressponding to point 0eigenprojection.It studies the relationship between two matrixes with the samecharacteristic eigenprojection of core inverse and gives the pertubation of the core inverse in the case of same eigenprojection.

The existence of periodic oscillationsfor a five-neuron recurrent neural network with time delay between neural interconnections is investigated.New criteria are proposed to guarantee the existenceof periodic oscillations for network model.Numerical simulation validates thismethod and demonstrates the correctness of theoretical analysis.

A subgroup H of a finite group G is calledπ-quasinormal in G if it is permutable with every Sylow subgroup of G.π-quasinormal subgroups are used to determine the structure of finite groups and obtain some sufficient conditions for a finite group tobe a p-solvable group or supersolvable group.

Counting matching problem is an intractable problem.Considering its corresponding parameterized problems p-deg-#MATCHING andp-#MATCHING,p-deg-#MATCHING is proved to be fixed-parameter tractable and p-#MATCHING has a randomized approximation scheme of fixed-parameter tractable.

Mahonia bealei fruit was extracted using ultrasonic wave to get extract.Based on the study of scavenging activities of extract against DPPH· and ABTS⁺·,the influence of solvent effect on scavenging activities was studied and the fluorescence-quenching spectroscopy was employed to study their chelation abilities with copper(Ⅱ) and ferrous ion.The results showedthat IC₅₀ of the extract of Mahonia bealei fruit scavenging against DPPH· and ABTS⁺· were much lower than 10 g/L,implying that the extract had goodscavenging activities,while the solvent effects had important influence on scavenging activities.In addition,the extract of Mahonia bealei fruit showedgood chelation abilities with copper(Ⅱ) and ferrous ion,with the quenching constants Ksv 7.57×10² and 3.09×10³ L·mol^-1,respectively.

The anti-HCMV effect and its potential mechanism ofthe water extract from Cordyceps sinensis is investigated.Cytopathic effect(CPE) method is utilized to examine its toxicity to human fore-skin fibroblasts (HFFs) and the anti-HCMV effect in vitro of the water extract.Moreover,western-bolt is used to investigate the mechanism of anti-HCMV effect of the water extract.The results show that the water extract can inhibit HCMV replication effectively at the concentration of 0.3 g·L^-1.In HCMV-infected cells,the expression of immediate-early protein IE1、IE2 and early protein UL84 can be inhibited to a large degree when treating with the water extract.

The mechanochemical decomposition of organic compound reactive turquoise blue KN-G was studied by using an oxide of NiO.The effects of grinding-time,material ratio,ball material ratio and rotation speed were also investigated.Thedecoloration rate and COD removal rate were measured.The products analysis was carried out with the use of X-ray diffraction detection (XRD) and IR.The results indicate that reactive turquoise blue KN-G was degraded effectively with mechanochemistry method.

Six species of Angiopteridaceae are reported for thefirst time from Guangxi.They are Angiopteris helferiana,A.caudatiformis,A.howii,A.wangii,Archangiopteris hokouensis,Ar.henryi.And a key to species of Angiopteridaceae from Guangxi is given.

Based on investigating data of 10 000 m² of 9 plots,details of phytocoenological characters of the community,such as species structure and community diversity,were provided.The main results are:①vascular plants in the original forest are very divers while neither species in genus or genusin families is scarce.Up to 61.74% of the total genera constitute only one species.Dominant species are unclear in tree layers,and the species with biggest important value may not be those in canopy.Shrub layer are dominant with species ofBambusoideae,Lauraceae,Fagaceae,Myrsinaceae,and herb layer are dominant with species of Urticaceae,Acanthaceae and Dennstaedtiaceae;②differ from the typical evergreen broad leave forest in the same belt,which shrub layer is the highest in all layers of the community,except species richness,Shannon-Wiener index,Simpson index and evenness index of tree layer are the highest,following by shrub layer and herb layer in turn;③The fact that the community has complicated speciescomposition but unclear dominant species may indicate that most of species havebeen survival in the long history and the forest contain much information of what it should have had.It may provide effective reference to afforestation to thelarge rocky desertification areas in similar karst environment in Southwestern China.

Lijiang National Aquatic Germplasm Resources Reservefor Spinibarbus hollandi and Sinocyclocheilus was examined and approved by Chinese Agriculture Ministry on December 22,2008.There was three natural resources investigations of aquatic organism in Lijiang,which were conducted during 1974—1976,1980—1982 and 2006—2008.The results showed that boththe species and quantity of fishes and the biodiversity of fishes were going through a degression.There has been no investigation of fish resources after the Reserve were established.In order to evaluate the effect of Reserve to the fish resources,an investigation was carried out for a whole year,thestatistical results showed that there has been a little restoration of fish resources in Lijiang.There were 74 species found in the Reserve,which belonged to 62Genus,17 Family,6 Order.The species of fishes were promoted from 66 in theinvestigation of 2006—2008 to 74,and the fish species diversity ofD_G-F were promoted from 0.443 to 0.596.All of these results showed that the establishment of the Reserve was very important in the protection and restoration of natural fish resources and other aquatic organisms.

A cave-dwelling loach,Oreonectes guananensis sp.nov.was recognized based on collections from several caves in Guan'an village,Changmei town,Huanjiang County,Guangxi,China,in June 2010.It can be distinguished from all known Oreonectes species by the following combination of characters:7 branched dorsal-fin rays,5 branched anal-fin rays,10～11 branched pectoral-fin rays,7～8 branched pelvic-fin rays and 13～17 branched caudal-fin rays;head compressed;eyes present;surface of lower lip covered with shallow longitudinal groove;dorsal-fin origin opposite or posterior to vertical line of pelvic-fin origin;caudal peduncle without caudal-adipose keels;edge of caudal fin rounded;scales present;cephalic lateral-line system developed;lateral line incomplete,with 7～13 pores;a dark brown stripe present along lateral line;posteriorchamber of air-bladder developed,extending to nearly pelvic-fin origin.

Some chemical parameters and phosphase activity of the overlying water in the centre of Huixian Karst Wetland were analysed by 4-time water sampling (0.5 m deep) at 7 districts (the central wetland was divided into 7 districts) (10～15 sampling points in every District) in early stage,metaphase,and late stage during the dry season from 2009 to 2010.Results of the researchsuggested that the overlying water body in Huixian Karst Wetland was in a eutrophic state of.Scale Ⅳ～Ⅴ,refered to the evaluation criteria of Taihu Lake water body.In this wetland,District Ⅲ and Ⅳ are the two largest ones and so their principal character is representative.The plants growing in these two districts dobetter in adsorption,intake or precipitation of phosphorus.But the phosphase activity of the water body in District Ⅲ is much higher than that in District Ⅳ,which shows that composition of flora species in District Ⅳ is not propitious toset free its internal phosphorus.Based on the differences in TP,SRP and phosphase activity between District Ⅲ and Ⅳ,it is suggested that the restoration of aquatic vegetation in Huixian Karst Wetland should adopt the pattern of DistricttⅣ:Cladium chinense (planted in the shore zone)-Vallisneria denseserrulata/Hydrilla verticillat/Potamogeton malaianus (in the ballow).The relationshipbetween phosphase activity and SRP is inverse proportion for the whole researched water bodies in Huixian Karst Wetland in early stage,metaphase of its dry season,which suggests that phosphase activity can work as one of the prewarning information for eutrophication tendency of the water body in this Wetland during the testing period.

Decision tree classification is an effective classification method in data mining,but its performance is severely affected by large dataset.This paper addresses the design and implementation of a parallel decisiontree classification algorithm based on MapReduce programming model.Experiment results show that this implementation works better than implementation based on other parallel programming models while running on more nodes.

An active learning method of pair-wise constraints based on error correction is proposed in this paper,and stopping criterion is also presented in order to get better clustering result with less pair-wise constraints.Experiments on the UCI benchmark datasets and artificial datasets show that theperformance of semi-supervised clustering algorithm with the proposed strategyis better than that of compared strategies.In addition,the clustering result of each tested dataset is acceptable under the stopping criterion.

Semi-supervised clustering is a new research direction of machine learning in recent years and an important branch of data mining,which has gradually become an useful tool in many areas.However,in the research for semi-supervised clustering now,especially when the number of classes in labled informations less than the entire data set,its clustering accuracy is not good.On the basis of the existing semi-clustering technology,the similarity of the samecluster of documents is improved by feature weighting with better clustering result.In order to verify the validity of this idea,experiment is carried out not only on the single-language data sets,but also the Sino-British data set in the labled document containing only Chinese or English language.The experimental results show that the method performs well.

The association rules between transactions are concentrated on meaningful association between different services in data mining.The existing algorithms can realize the largest conversion of frequent itemsets amongand within affairs,and find the relationship between different users.However,indealing with the affairs of the larger database,a large number of false rules are generated due to the correlation analysis is performed on the original database.This paper proposes a mining algorithm of association rules for inter-transaction among affairs based on cluster analysis.Through simplifying the complex dataset by cluster analysis,the proposed approach can eliminate redundant data,reduse the size of dataset and avoid the need for multipole scanning of the database and the generation of a large number of false rules.Experimental results indicate that the proposed method has higher efficiency and the ability of more precisely predicting the interests of users

The metadata is a key issue of neuroimaging databaseresearches.However,the existing metadata organizational forms of neuroimaging data are too simplex to effectively support the systematic BI research oriented analytical queries for data information.The construction of metadata cubes can solve this problem.In this paper,a Data-Brain based approach is proposed to construct the metadata cubes of neuroimaging data.Based on the existing technologies related to data cube constructions,a type of novel metadata cube of neuroimaging data,i.e.,Provenance Cube,is defined.A Data-Brain based semi-automatic approach is also designed for Provenance Cube constructions.The construction instancestated in this paper shows that different topics oriented metadata cubes of neuroimaging data can be constructed quickly to support various analytical queries of data information coming from different aspects of systematic BI research.

Phrase dictionary used in search engine can effectively improve and enhance the performance of information retrieval.“V+N₁+N₂”structure is a common phrase structure in Modern Chinese.This paper,based on Sogou log corpus,analyzes the “V+N₁+N₂” verb and the noun phrases and presents the processes of relational data from the perspective of the characteristics of constructional elements,semantic features and syntactic function.In terms of not changing the original semantic based on different semantic relations,thisarticle proposes a corresponding transformation model and provides a further theoretical support for phrase dictionary building of future search engines.

“N₁+N₂+V” structure is a basic form in composing noun phrase.Based on Sogou log corpus,this paper focuses on studying the nounphrase of “N₁+N₂+V” structure in search engine query logs.According to thefeatures of corpus,the phrase is studied from four aspects:syntactic characteristic,phrase-level segmentation and resolution,syntactic function and semanticrelationship,focusing on analysis of the semantic relationship of N₁ and N₂,N₂and V.Sogou query logs are deeply analyzed and confirmed,which provides a theoretical basis for the construction of phrase dictionary used by search engines.

Projection pursuit is used to find the projection direction which reflect the intrinsic structure and features of the originaldata.So the high dimensional data can be studied and analyzed through lower dimensional space.For the curse of dimensionality in text classification,the multi-dimensional data is turned into low-dimensional space by using projection pursuitmodel.The nature of the problem is that it is difficult to determine thedirection of projection pursuit and the calculation is not easy,especially whenthe projection direction has more indicators.This paper proposed the projection pursuit model using immune evolution algorithm,which can effectively solve theoptimization problem of the projection direction.In the experiment,it's appliedto text classification including the Reuters-21578 document sets and FuDan documentsets.The experimental results show that this method can not only solve the curseof dimensionality in text classification,but also obtain a satisfactory classification results.

Text classification is one of the key techniques in text information process,which includes how to establish vector model,select feature and train classifier.EIBA and DHCHi2 are integrated to select the features of text and the features are used as parameters in a fuzzy soft set theory.Then a new technique of text classification is established based on a fuzzy softset.Experiments show that the technique is effective,and the ratios of accuracy andrecall are improved comparing with other methods.

A content extraction method based on extended label tree is proposed.Web page cleaning and auxiliary information for extracting purpose are realized,and the coordinates of position are also set during the construction phase of extended label tree.Text nodes are regarded as the identifiers ofthe content region,then,the neighbor text node set with maximum coverage is selected and revised to form the final content region.Through the neighbor first traversal algorithm,the title node is located and additional properties are extracted.Experimental results show that the proposed method can achieve high-precision for common article page extraction and has good adaptability.

Most previous works on Web information extraction seldom use associations among Web page blocks.In order to solve this problem,this paper proposes an automatic Web news content extraction approach based on conditional random fields (CRFs).Firstly,it parses a target news page to a DOM tree.After eliminating invalid nodes,pruning subtrees and deleting single nodes in the tree,it uses heuristic rules to segment the DOM tree to blocks and converts theseblocks into a data sequence.Then,it defines feature functions to extract each block's own state features and neighbor blocks' category transition features.Finally,by labeling the data sequence based on CRFs,it identifies each block's category to extract the page's content.Experimental results indicate that this approach is precise and adaptable for Web news content extraction,and importing associations among page blocks can improve Web news content extraction.

The traditional text feature selecting and weightingmethods don't take into consideration the semantic relationship between terms,sothat the weight of features cannot be reflected.In this paper,one weighting methodof ontology-based feature is presented,which solves the problems mentioned above.By mapping feature item to concept and then adjusting the weight of the feature item,the selected feature can better reflect the text characteristics.At last,to illustrate the effectiveness of the method,the paper introduces an example,which shows that it can improve the accuracy of feature extraction to a certain extent.

Aiming at the current problems of semantic inaccuracyand horizontal summarization existing in multi-document summarization systemsaccording to the similarity of theme,the paper presents a mo-del of multi-document summarization based on HNC,thus it can convey follow-up report information of articles of same event (i.e.Vertical Summarization) with short text.The experiment shows that the method exceeds the TF×IDF method in coverage rate and redundancy rate of information and summarization fluency.

Opinion analysis based on opinion-object extractionaims at mining opinions from multi-field reviews.This technique analyzes opinion reviews by extracting opinion objects and grouping the opinion objects automatically.A ChunkCRF model is constructed to divide the opinionatedsentence into particular chunks which helps to identify the opinion objects andother opinion elements effectively.In the case of multi-objects in one sentence,syntactic structure is analyzed for a decision.In the case of implicit opinionobjects,opinion object is generated by a semantic method.On this basis,opinion analysis based on opinion-object extraction is practiced.

Expert Entity Homepage Recognition is one of the keypoints in expert search.In this paper,a method based on J48 is proposed.2113 Chinese expert entities and the corresponding entity homepages are collectedby analyzing the expert resources,and the expert entity features relatedto the features of link and webpage content are defined.Besides,these features are also extractedto form a training data set;and then different learning algorithms with different features are adopted to recognize the expert homepage for finding the most effectiveclassification features and homepage recognition learning algorithm.The experiment results show that the best method is obtained by using J48 algorithm,specifically,when the features of link and webpage content are combined with,theexpert homepage recognition accuracy rate reaches 81.05%.

To expediently and fast utilize the information of Deep Web,content acquisition of website databases is an effective method.This paper introduces the research on acquisition of Deep Web content based on domain-Ontology.Using training database to train the domain-Ontology construction,itobtains Deep Web content after querying database by keywords through the effectible concept of domain-Ontology to obtain query keywords.The experiments verify that this method can obtain most of Deep Web content effectively,andusers can more expediently and faster utilize the information of Deep Web through search engines.

The paper introduces the recognizing and extracting methods of mathematics expressions in formula-based mathematics search engine.Itsummarizes the corresponding features of MathML,OpenMath,LaTex and Infix when they are embedded in a Web page.A feature-based heuristic method of recognizing and extracting mathematical expressions is given in the paper.The experimentsproves that the method is effective and useful.

Collaborative filtering has been widely used in E-Commerce recommendation systems,but the sparsity of data affects the quality of collaborative filtering recommendation.A two-stage collaborative filtering algorithm is proposed based on random walking and cluster-based smoothing.For off-line stage,calculate the correlation betweenitems,suggest anew method which describes the correlation between items by cumulating weightedtransition probability of each step.Cluster items according to the item correlation matrix,then smooth the unrated data by using clustering information.For on-linestage,search the target item's neighbors according to the correlation between items cumulated during the off-line and predict.This method can enhance the description of the correlation between items.The experiment results illustrate that searching neighbors according to the item correlation matrix will become more accurate,which can effectively relieve theimpact of sparse data and improve the quality of recommendation.

For dealing with the imperfection of the guidance systems for employment in universities,a job recommendation system is designed anddeveloped in this paper.By comparing the essential features between current and past graduates,the similarity can be calculated according to therelevance between graduates and enterprises.Using random walk algorithm,the index of being employed can be obtained.At last,according to the variables calculated above,the system can recommend suitable jobs for different graduates by putting enterprises in optimal order.Experimental results show that the system fits the purpose of design.

In order to solve the problem the inaccuray of texture image retrieval the traditional methods,a method of multi-level image texture recognition based on tolerance granular space is proposed.Considering the loss of color in texture image,a color matching method is addressed based on granularity level.Combining the above two methods,an improved similarity algorithm which integrates color and texture features is given.Comparative experiments showed that the method is effective and reasonable.