C. Li, B. Stratman, and S. Mahadevan (USA)
Machine learning, Classification, Imbalanced data, Decision tree, Support Vector Machine, Wheel inspection
Railroad wheel inspection attempts to identify failing wheels from a large population of wheels in service. This is a critical yet time consuming task. This paper presents a machine learning approach to automate the identification process using collected data from wheel inspection. Decision tree based and support vector machine based classification methods have been applied to the wheel inspection data analysis. A variation of the bagging ensemble approach is developed to improve the classification accuracy. The results of these methods achieve an identification accuracy of 80%. Analysis of the rules and models derived, as well as comparisons of the classification results obtained using the two base classification approaches are presented.
Important Links:
Go Back
