In view of the inadequacy of K nearest neighborhood (KNN) algorithm in text-processing environment in vector space models,this paper puts forward an improved KNN method of text categorization in accordance with self-organization mapping neutral network theory(SOM),feature selection theory and pattern aggregation theory.

In this paper, three methods for text feature dimensions reduction are presented: The first method reduces the dimensions based on pattern aggregation theory and an improvedχ~2 statistic, and then the better accuracy of categorization is acquired;

The method firstly reducestext dimension with Pattern Aggregation theory that uses class label, then makes thetext dimension further lower by LSI method.

The results indicated that the spatial distribution pattern and aggregation intensity were different in different environmental conditions,while the tendency of pattern aggregation was generally parallel. The figure of pattern scale and pattern intensity showed that plot 2 clumped in 25 and 100 m2,and plot 3 clumped in 150 m2,while plot 1 performed the pattern of random distribution in all quadrat scale.

In most of the reported studies on spatial pattern of insect populations,the natural habitat unit (NHU) of population individuals is essentially assumed as point,i.e., the number of individuals within a NHU as basic unit for surveying,assessing and interpreting spatial pattern, But in fact the NHU of insect population is generally a multidimensional structure rather than point. Therefore the much spatial information may be lost owing to this inappropriate assumption. In this paper, pine tree,the NHU of Dendrolimus...

In most of the reported studies on spatial pattern of insect populations,the natural habitat unit (NHU) of population individuals is essentially assumed as point,i.e., the number of individuals within a NHU as basic unit for surveying,assessing and interpreting spatial pattern, But in fact the NHU of insect population is generally a multidimensional structure rather than point. Therefore the much spatial information may be lost owing to this inappropriate assumption. In this paper, pine tree,the NHU of Dendrolimus tabulaeformis,is considered as one-dimensional axis composed of strata of verticillate-branches(SVBs) rather than points. Based on the above idea,spatial patterns of Dendrollmus tabulaeformis larvae and pupae on every SVB were assessed and interpreted with Taylor's power law model and the author's reinterpreted Taylor's power law,and the law of spatial patterns varying among strata was analysed with Fuzzy Clustering Analysis (FCA),Grey Clustering Analysis (GCA) and Trend Curved Surface Analysis (TCSA)methods. Also within-pine-tree vertical distribution aggregation of population was discussed. It coneiudes that: (1)according to the author's reinterpreted Taylor's power law, the spatial pattern of larvae and pupae on every SVB follows inverse densitydependent aggregation,and the critical density m of the population aggregation was calculated, so the spatial pattern continuum on every SVB can be deseribed quantitatively; (2)the resuits of FCA,GCA,and TCSA show that the alw of spatial pattern aggregation varying among Strata is non-liner and oscillatory,and the Strata can be clustered into several types according to the characteristics of the population aggregation; (3)within-pine-tree vertical distribution also follows inverse density-dependent aggregation.

The module morphological characteristics of Leymus chinensis, Phragmites communis and Kalimeris integrifolia were measured and analyzed by sampling randomly in thirty sites. To explore the mechanism of competition among main plant species on grasslands in Changling County. Fifty L.chinensis plants in each L.chinensis, L.chinensis + Ph.communis and L. chinensis + K.integrifolia community, twenty K.integrifolia plants in each K.integrifolia, L.chinensis+ K.integrifolia and Ph.communis + K.integrifolia...

The module morphological characteristics of Leymus chinensis, Phragmites communis and Kalimeris integrifolia were measured and analyzed by sampling randomly in thirty sites. To explore the mechanism of competition among main plant species on grasslands in Changling County. Fifty L.chinensis plants in each L.chinensis, L.chinensis + Ph.communis and L. chinensis + K.integrifolia community, twenty K.integrifolia plants in each K.integrifolia, L.chinensis+ K.integrifolia and Ph.communis + K.integrifolia community were taken respectively. Twenty Ph.communis plants in each Ph.communis, L.chinensis+Ph.communis and K.integrifolia +Ph.communis were sampled. The individual height, leave length, leave width, leave number and internode's distance of L.chinensis, Ph.communis were measured respectively, as well as the individual height, shoot length, shoot number and shoot distribution pattern (aggregation distribution, regular distribution) for K.integrifolia. The data on plant characteristics were analyzed by SPSS package. The experimental results showed that the module morphological characteristics of L. chinensis, Ph.communis and K.integrifolia changed under the interaction among coexisting plants and the hierarchy of competition for light were formed as followed: Ph.communis>K.integrifolia>L.chinensis. However, the hierarchy did not mean that L.chinensis had disadvantage in competition. The trade-off between adaptation and interaction were achieved by the vegetative propagation. The plasticity of plant module morphological characteristics was affected by the light resource, the interactive relationship and the adaptation to environment.

In view of the inadequacy of K nearest neighborhood (KNN) algorithm in text-processing environment in vector space models,this paper puts forward an improved KNN method of text categorization in accordance with self-organization mapping neutral network theory(SOM),feature selection theory and pattern aggregation theory.This paper employs feature selection theory and pattern aggregation theory to reduce feature space dimension.And because each dimension of VSM models possesses the same weight,which...

In view of the inadequacy of K nearest neighborhood (KNN) algorithm in text-processing environment in vector space models,this paper puts forward an improved KNN method of text categorization in accordance with self-organization mapping neutral network theory(SOM),feature selection theory and pattern aggregation theory.This paper employs feature selection theory and pattern aggregation theory to reduce feature space dimension.And because each dimension of VSM models possesses the same weight,which is not suitable for text-processing environment,this paper suggests applying SOM neutral network to calculate the weight of each dimension of VSM models.Combining the two improvements,this paper efficiently reduces the dimensions of vector space and raises accuracy and speed of text categorization.