Leveraging Machine Learning of Indian Railways Public Procurement Data for Managerial Insights

This paper proposes and demonstrates (a) a supervised machine learning methodology to predict the expenditure incurred on procurement, repair, and reconditioning of components as well as expenditure incurred on procurement of fuel based on the performance of the Indian Railways and (b) an unsupervised machine learning methodology to classify the good and poor performing administrative zones, using the data of expenditure incurred on procurement, repair, and reconditioning of components as well as expenditure incurred on procurement of fuel and the performance. The first methodology will aid managers in determining whether the expenditure incurred is more than what should be incurred. Further, it may also benefit the managers in fine-tuning the frequency of replacement of components. The second methodology will assist managers in searching for best practices of maintenance in good-performing zones, which can be propagated in the poor-performing zones to lower the overall expenditure on maintenance.