A practical approach to variable selection methods - A case study at Volvo Construction Equipment.

Details

Supervisors

Saerens, Marco

Faculty

Louvain School of Management

Degree label

Master [120] en Ingénieur de gestion

Abstract

Many challenges related to the emergence of new automation techniques and their exponential growth will have to be addressed in the future. Indeed, traditional statistical models developed in the pre-Big Data/High Dimensional datasets era are no longer fit to process current company data sets. Therefore, variable selection methods and algorithms were created. The purpose of this research paper will be to develop a practical case study at Volvo Construction Equipment which consists to compare two feature selection techniques namely, the forward-stepwise selection method and the Lasso method, in order to identify which one gives the best result in terms of interpretation. The theoretical framework developed in the first chapter focuses on the theoretical background and retrace the evolution of the way statistics were used and should be adapted considering the current fast changing environment. In addition, definition of all relevant statistics concepts directly related to the empirical study are provided. The empirical analysis is developed in the second chapter. It first provides an in-depth introduction of Volvo CE; starting with their market positioning and the presentation of their processes, from data collection to the practical processing of the selected statistical models. The rationales behind the choice made by Volvo regarding both, feature selection method and statistical models, are also explained. Eventually, even considering several limitations presented below, we will see how the Lasso method provides an improvement over the forward stepwise selection at least in the usage context of Volvo CE using German total demand for industrial equipment.