Approach to Detection and Elimination of Contradictions in big Data Bases for Analytics Concerns of Urban Passenger Transport

The aim of the study is to develop an approach for detecting and eliminating contradictions in big data bases for analytics concerns of urban passenger transport. The relevance of identifying and eliminating contradictions in big traffic data flows is due to the fact, that they are converted into significant information to use in management tasks. A feature of the proposed approach is the implementation of a universal module for two-level contradictions detection in the selected data set, using statistical methods and machine learning algorithms. The developed module «Detecting and eliminating inconsistencies» includes three methods of detecting and eliminating conflicting data: 1) methods of statistical processing, 2) Grubbs criterion; 3) machine learning algorithms: k- mean and multilayer neural network. The software implementation of the system for the proposed approach to identify and automatically eliminate contradictions is used to solve the formation of the city bus routes schedule. The article presents the stages and results of numerical experiments for the selected data set in the developed system for the proposed approach to identify and automatically eliminate contradictions, when compiling and adjusting the schedule of bus routes in Almaty city.