Deep Learning and Hybrid-Combinatorial Models for Telecom Sales Forecasting : Can Neural Networks Outperform Autoregressive Models ?

Details

Supervisors

Candelon, Bertrand

Faculty

Louvain School of Management

Degree label

Master [120] : ingénieur de gestion, à finalité spécialisée

Abstract

The telecommunications sector has become a dynamic battleground where rapid technological advancements, evolving consumer behaviors, and intensified competition define success. In Belgium, Proximus stands as a leader in this industry, yet it faces a critical challenge: understanding and optimizing the interplay between lead sales (driven directly by marketing actions) and non-lead sales (sales occurring without direct marketing influence). This duality not only impacts sales forecasting but also the efficacy of multi-channel marketing strategies. This thesis investigates the capability of modern forecasting methodologies, particularly deep learning models and hybrid approaches, to outperform traditional auto-regressive techniques like ARIMA and SARIMA. It further explores how marketing dynamics and economic indicators influence sales patterns, aiming to identify key drivers and enhance predictive accuracy.