Data Exploration Methods for Transport System Dependability Analysis

This work is devoted to application of data exploration and data mining techniques for analysis of monitoring databases of a large transport system. The analyses discussed focus on discovering relationships between key metrics of a transport system as such availability / usage profiles of the fleet and various factors on which they apparently depend (such as age, etc.). We demonstrate that building an OLAP system on top the monitoring / maintenance database opens new possibilities for transport system owners to efficiently discover such relationships on their own. This approach turns their monitoring database into a decision support resource in such areas as: optimization of maintenance policies of the transport fleet or discovery of untypical patterns in data (e.g. fraud related). The concepts discussed are illustrated by a number of examples based on real data from the transport system of the Polish Post.