Forecasting methods applied to industry
Tutor / director / evaluatorMorgenthaler, Stephan
Document typeMaster thesis (pre-Bologna period)
Rights accessRestricted access - confidentiality agreement
In industry, forecasting future needs is a crucial step in planning processes. Good forecast will allow the company to be in a better position when negotiating prices with suppliers, avoid stockruptures, detect trends, analyse the demand, avoid not having enough staff to run the business, etc. Different forecasting techniques have found many different applications in a wide spectrum of fields: ‐ Mistake detection and error correction ‐ Cluster products/countries/months based on similar behaviours ‐ Clean‐up databases Statistics provide companies a big set of tools which allow constructing such models. Moreover, it is very important for a company to have all these techniques explained in a rigorous but clear way in order to use the correct method in the appropriate situation. The aim of this project is to create a statistic toolbox for companies wanting to use rigorous statistics in their business when analysing data and not getting lost in theory. The methods which are going to be discussed will be the following: ‐ Multiple regression models: ANOVA regression models and the particular case of missing information and components of variance. ‐ Time series: Non‐seasonal and seasonal regression models, ARMA and ARIMA models ‐ Neural networks This Project has been done as a part of a six‐month internship in Procter & Gamble, (Geneva, Switzerland), supervised by Lieven Tijtgat. Besides, it has also been presented as a Master Thesis for the Master in Mathematical Engineering at EPFL (École Polytechnique Fédérale de Lausanne), supervised by Stephan Morgenthaler.
ProvenanceAquest document conté originàriament altre material i/o programari no inclòs en aquest lloc web