Data Driven Detection of Railway Point Machines Failures

Abstract

In this paper, a novel approach to early detection of railway point machines failures is presented. Easily accessible data from Centralized Traffic Control (CTC) systems, along with meteorological data, are utilized to build a classification system recognizing risk factors for railway point machine failure. We present and discuss a framework that aims at extracting information from the raw railway logs, and discuss the issues that need to be solved to make the framework properly operational. We show that ensemble methods utilizing decision trees are able to provide meaningful classification accuracy for this problem.