Classification and regression models to predict the cost to serve new customers in the industrial gas business

Identifying the cost-to-serve a customer is one of the most challenging problems in Supply Chain Management because each customer has his own characteristics (large/small orders, frequent/less frequent orders, etc.) and thus engenders a different logistics cost.For the particular case of the industrial gas business, we are interested in predicting the cost to deliver bulk (liquefied) gas to a new customer using a linear regression model. A multifactor regression model is built based on similarities between customers. Prior to the regression analysis, a data classification technique is used to cluster customers who exhibit the same characteristics. Groups of customers are represented by hyper-boxes and a linear regression is developed for each box. Instead of having a single model for all customers, the idea is to build a linear piecewise regression model to increase the accuracy of the prediction. For data classification purposes, we developed two new Mixed Integer Linear Programming (MILP) models which can be used on any data set in which the classification attribute has a continuous value. A sensitivity analysis is carried out through the study for adequacy testing purposes.Key words: Cost-to-Serve, Data Classification, Hyper-box, MILP, Regression Analysis, Industrial Gas Business...