logo SBA

ETD

Archivio digitale delle tesi discusse presso l’Università di Pisa

Tesi etd-04082013-142202


Tipo di tesi
Tesi di laurea magistrale
Autore
ORSINI, FABIO
URN
etd-04082013-142202
Titolo
Development of artificial neuronal networks as prediction instrument.
Dipartimento
INGEGNERIA DELL'ENERGIA, DEI SISTEMI, DEL TERRITORIO E DELLE COSTRUZIONI
Corso di studi
INGEGNERIA GESTIONALE
Relatori
relatore Prof. Dini, Gino
Parole chiave
  • Artificial Neural Network
  • Forecasting
  • Time Dynamic cost
Data inizio appello
24/04/2013
Consultabilità
Non consultabile
Data di rilascio
24/04/2053
Riassunto
This work deals with the development and validation of an Artificial Neuron Network as a prediction tool for time dynamic costs.
Time dynamic costs are variables characterized by important variations within a relative small timeframe.
These type of costs are the major cause of unexpected fluctuations of the total product cost.
The pricing calculations are usually prepared in the early stages of processing a customer’s inquiry, thus an unexpected fluctuation of the costs may cause an underestimate or overestimate of the cost and a loss of money and competitiveness for the company.
To solve this problem a solution could be the utilization of the artificial neural networks which usage in the field of economics is still underrepresented.
The first part of the work describes the state of the art of the methods of cost estimation and prediction models, highlighting the main aspects and the limitations.
The state of the art is concluded with a deeper overview of the theory about the artificial neural networks.
To prove the efficiency of this tool two different ANN are developed basing on two different approaches for the selection of the input, to predict the price of the copper as example of a time dynamic cost.
The first approach is based on historical data and the second on some parameters that can have some impact on the future final price of the material.
Therefore, in the second part of the thesis all the phases to develop and validate the neural networks are described in detail, with a conclusion about the advantages and limitation of each approach used.
File