Advanced predictive models for large scale (big) data analytics

Details

Supervisors

Catanzaro, Daniele

Faculty

Louvain School of Management

Degree label

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

Abstract

The aim of this thesis is to create a forecasting system for the return of tire carcass for a famous tire producer by considering the consequences of the COVID-19 outbreak. This system will advise the manager in the sourcing of raw materials process. By providing more accurate estimations of the demand, the company could reduce its operating costs by avoiding waste and the cost of opportunity of being out-of-stock. CRISP-DM methodology will be used throughout the project. This method involves an analysis of the business and data but also the pre-processing phase which will prepare data to implementation of the models. The problem is tackled by using an extrapolation and a causal approach. Therefore, ARIMA model and LSTM neural network are used for the extrapolation approach while embedded multiperceptron neural network is applied for the causal approach. This thesis highlights the accuracy performance of neural networks in a complex and non-linear environment using TensorFlow Python library. To provide an automate and scalable result, the database management system Microsoft SQL Server manages the pre-processing and the deployment process of the causal approach. Recommendations are also made to help the company to be more data driven and hence improve the accuracy performance of the deployed forecasting systems.