Research And Application Of Land Disaster Early Warning Model Based On Data Mining

China is a country where geological hazards are relatively frequent,and each occurrence of a geological disaster damages the lives and property of the public to varying degrees,and hinders the economic development of society.Therefore,effective early warning of geohazards has become a top priority in managing disasters in China.This thesis combines data mining technology to establish a geohazard early warning model with correlations,and applies the early warning model to a geohazard monitoring system for the purpose of achieving higher early warning accuracy.The main contents and results of the research in this thesis are as follows.(1)Based on the traditional FUP algorithm,the classical incremental association rule algorithm is improved: Since the classical incremental association rule mining algorithm in data mining techniques,the FUP algorithm,has two disadvantages,such as although the computation of the transaction itemset part can be reduced by using the original association rule mining information,it still requires frequent scanning of the transaction database,which makes the computation increase;using the string comparison function to complete the statistical The use of string comparison functions to complete the statistical item set support is time consuming,etc.Therefore,an incremental association rule mining algorithm based on sorted compression matrix FBSCM(FUP Algorithm based on the Sort Compress Matrix)is proposed in this thesis to overcome the above drawbacks.In this algorithm,1)the transaction data set is mapped into a Boolean matrix and the storage of the matrix is changed;2)the invalid rows and columns in the matrix are compressed using the correlation property;3)the items of the matrix are also incrementally sorted according to the idea of item set orderliness.Combining the above three points improves the overall mining efficiency of the algorithm.(2)Multi-dimensional experimental comparison based on the improved algorithm to test the effectiveness of the improved algorithm: The single variable method was used as the experimental method to compare several different algorithms by adding different numbers of incremental datasets,changing the minimum support threshold and selecting datasets with different data characteristics.The experimental results consistently show that the improved algorithm in this thesis has higher time performance compared to the traditional FUP algorithm,especially when the amount of incremental transactions is larger or the minimum support is smaller.(3)Development of a data mining based early warning model for ground hazards: Firstly,the multiple monitoring objects(cracks,dips,surface displacements and rainfall)to be studied are selected and the data are investigated and analysed to extract data features;then appropriate data pre-processing is carried out according to the data features;then the improved FBSCM algorithm is used to obtain the correlation relationships between multiple monitoring objects and strong correlation rules are extracted for further analysis,and finally an early warning model with correlation relationships is built.(4)The application of early warning models for geohazard monitoring systems: The geohazard early warning model established based on data mining technology is applied to the geohazard monitoring system,and the early warning model is used to conduct relevant early warning experiments on the input monitoring objects,and the experimental results are compared with the geohazard early warning model based on the traditional single-parameter threshold judgment early warning.The results show that the geohazard early warning model established in this thesis can better improve the effectiveness of early warning information,reduce the number of false warnings and improve the reliability of early warnings.