• architecture
• arima
• automotive
• azure
• cloud
• exponential smoothing
• holt-winters
• lstm
• sales forecasting
• time series

The main objective of this thesis was to develop a tool that could be able to predict how many units of product will be sold globally in the coming months and years, divided by country and region.

The challenge regarding this goal lies in the need to accurately forecast every type of product manufactured by the multinational client company (belonging to the automotive industry) in every country where it is currently operating. Because of the sheer amount of predictions to produce, the problem quickly became computationally demanding for conventional systems: the tool designed and developed during this project was built leveraging cloud capabilities provided by Microsoft Azure.

Finally, after a testing and validation phase characterized by a very close collaboration with the business users employed at the client firm, the tool was successfully deployed and integrated in the client's decision making processes.