Study Of Site Knowledge Discovery Based On Multivariate Forestry Information

As the external environment of forest growth, Site directly affects forest productivity and health status. With the development of forest resources survey and data acquisition technology, forest resource data already has a diversified, sea quantization and information characteristics. More abundant forest resources data implies a large number practical information and knowledge of the forest site. But for a long time, a large number site research (site productivity, site factors) still use traditional forest resources survey data, a wide range of forest resource data have not been integrated and fully utilized, and research tools rather monotonous, lack comparison research of variety of methods. This makes the massive forest resources data facing an increasingly significant problem of information explosion but lack the knowledge. In order to take full use of the diversified sea quantified forest resource data, found implied information and knowledge, to provide new method, theory and reference for site research. Use forest resources survey data, multi-spectral remote sensing data, digital elevation model data (DEM), based spatial data mining technology to extract a wide range of forestry information, use GIS as platform achieved multi-forestry information fusion and visualization. Using spatial analysis, decision tree classification, neural network prediction, space discrete field analysis of a series of data mining theory and methods, extracted and analyzed environmental factors on forest site productivity knowledge, factors related knowledge. The specific contents are as follows:(1) Site spatial data miningThe study based on DEM data and multi-spectral remote sensing image of the Neimengu Wangyedian forest farm. Using Clustering, aggregation, information composite, tracking and window spectral bands combination of spatial data mining technology. For forest site micro-topographic factors, macro topographic factors and biologicalfactor extracted and analyzed to obtain elevation, slope, aspect, slope variability, aspect variability, surface roughness, surface waviness, sloping, DVI, RVI, NDVI, TSAVI12site environmental factors. The research results enrich the site types of environmental factors, provide effective approaches for low-cost large area of forest site data mining researc.(2) Site barren performance predictionBased on the Micro-landform factor, macro-terrain factor and bio-factor extracted in the above part, binding the information of forest resources survey data constitutes multivariate forestry information. This article uses decision tree technology and artificial neural network technology analysis the relationship between the number of site factors and site barren performance. According to the structural characteristics of the multivariateforestry information, research using Boosting, Cross-validation, Group Symbolics, as well as local pruning mechanism improved C5.0decision tree, use the mechanism of the sensitivity analysis improved BP neural network model. Established the site barren performance prediction model, form8different experimental programs. The results show that the proposed two improved algorithms most suitable site barren performance prediction, the prediction performance is ideal. The study also found that number of environmental factors extracted by spatial data mining technology can expand the amount of site factors information, improve the prediction accuracy of the site barren performance, and provide potential for wide range of multi-temporal prediction.(3) Correlation knowledge extraction of site discrete space fieldSelect10typical discrete data in forest resources sublot survey data combined with small class spatial location attributes formed forest site space discrete field. Use information entropy theory, by icalculating the local space and local space and the overall coordination of quantitative analysis and extracted a number of discrete factors and site quality rating index. The results showed that different site types and dominant tree species the highest degree of correlation between site quality in10factors, and origin of forest showed independent relationship. Research overcomed inadequate that conventional statistical principles and gray system theory can not calculated site space discrete field, Achieved calculation and expression of site discrete factors.