Research On Spatial Data Mining Based On The Basic Of Rough Sets Theory

Our country is a landslide disaster-prone country.The detailed surveys ofgeologic hazard and the research of ecologic hazard has accumulated a lot ofhistorical landslides and related data during these years,but it is hardly to geteffective knowledge or rules from these history data.This resulted in “rich data butpoor information”.So it is very urgent and necessary to conclude the effectiveknowledge or rules from historical geologic hazard data to prevent and alleviate thegeologic hazard and forecast the disaster, so as to sublime from data to knowledge.In the prediction of the landslide of area,the index system to bescientific andreasonable is a very difficult issue.The current selection and classification of landslidefactor mainly rely on the qualitative guidance and quantitative method on thejudgment of experience and the classification of experts,it has strong subjectivity andthere is no established organic link between the factors, so it is easy to lead to largedata redundancy and it is not suitable for accurate computer simulation, themonitoring and forecasting the comprehensive analysis of large area.Rough set theoryis a depiction of incomplete and uncertain mathematical tools,it can effectivelyanalyze and deal with imprecise,inconsistent and incomplete information and discoverimplicit knowledge to reveal the potential law which could be used on the study ofthis problem.Thispaper, which selected Bei’chuan as the studing area, calculatingsensitivity coefficient of each interval of impact factor of landslide and classify thefactor of landslide based on sensitivity coefficient.Evaluating the importance of the impact factor of landslides and calculating its weigh and discovering decision rulesbased on rough sets theory. It is a try and exploration that applicate the rough setstheory on the field of geological disaster, briefly the major study of this paper andresults as below:(1)Summarized the system and methods of spatial data mining based on previousstudies, and simply introduced rough sets theory and basic knowledge.(2)Analysed the distribution and characteristic of the landslide which occuredbefore the5.12earthquake of Bei’chuan and the impact factor of landslides by usingpowerful spatial analysis of GIS such as spatial statistics, overlay analysis, bufferanalysis et al. Results indicates thatthe distribution of landslides which occurredbefore the5.12Earthquake of Beichuanon administrative region and aspect is uniform,more distribution on the elevation of800-1600m and slopes within20°-45°,along theriver and road,the landslidesis linear distribution and the number of landslides reducewith the increase distance,more distribution on the woodlands hilly dryland and thestrata of spy organization on Mao County group of the Silurian, more distribution onthe area of average annual rainfall from1340to1440mm.(3)Selected elevation, slope, affected distance of the river, the influence distanceof excavated road, average of annual rainfall, rock groups, faults, land as the impactfactors of the landslide occurred before the5.12earthquake of Bei’chuan, So as toavoid subjective classification by human, make classification of the impact factormore effective and rational, calculate sensitivity coefficient of each interval of impactfactor of landslide and grade the factor of landslide based on sensitivity coefficient.(4) Divided the gird unit of Bei’chuan and assigned values to the gird based ondivision table of state on impact factor of landslide,built two-dimensional decisioninformation table of landslide based on the properties table of the gird,analysis andmining by the two-dimensional decision information table of landslide based on roughsets theory, including approximate analysis of two-dimensional decision informationtable of landslide, calculation of core, reduction of attribute, calculating importanceand weighting of each condition attribute, export the decision rule of landslide andnon-landslide. Results indicates that eight selected condition attribute factor (impactfactor of landslide)for decision attribute (landslide and non-landslide)are the nessaryfactors and the core of decision attribute. The importance and weight of eigthcondition attribute is that land(0.252), average of annual rainfall(0.229),elevation(0.141), slope(0.123), affected distance of the river(0.109), rockgroups(0.072), the influence distance of excavated road(0.058), faults(0.015).The higher frequency of matching rules:rule90,103,107,108,129,142,126,155has a certainvalue.