Mapping Algorithm Research On 'Internet Plus' Civil Aviation Auto-service Master Data

With the rising degree of civil aviation information,the ‘Internet plus' category has been greatly touched by civil aviation business.But there isn't a complete master data management system at civil aviation information system.The master data which is heterogeneous is distributed in civil aviation subsystem because of the different naming nomenclature.To improve the quality of civil aviation master data and increase its value,a method which managing and mapping civil aviation data and ‘Internet' data should be researched in order to build a data platform and associate ‘Internet plus ' data.Firstly,candidate civil aviation data is selected according to the content of the civil aviation master data management.the master data identification method based on the analytic hierarchy process is used to identify the civil aviation master data.Secondly combined the features of data structure,proposes a method based on data contents which combined sparse auto-encoder and support vector machine(SVM)for data classification and mapping.This method maps the civil aviation master data to high-dimension space;Then the deep features were extracted by sparse auto-encoder neural network with sparsity constraint to maintain its sparsity;And the classification of single record was generated by the support vector machine.At last the classification of every column of data was determined by voting.The experimental results of civil aviation data indicates that the average accuracy of single record is 97.44% and the accuracy of a column of data is beyond 99.99% along with the raising automatization-level of heterogeneous civil data.Tertiary,a method of ‘Internet' data classify mapping based on address master data has been proposed which combined the features that the ‘Internet' data are all related to address data.The 4-level address tree has been set up in order to describe ‘Internet' data with standard address information.The unified management model of ‘Internet' data has been provided so the data structure is standard.And a Chinese address normalization matching method combining string,pinyin coding and five-stroke similarity is proposed.Experiments show that the matching accuracy rate of this method can reach 69%.Compared with using only the string similarity,the matching accuracy rate is higher.