Sistema ETD

Archivio digitale delle tesi discusse presso l'Università di Pisa

 

Tesi etd-04102013-100326


Tipo di tesi
Tesi di laurea specialistica
Autore
GLIOSCA, DIEGO
URN
etd-04102013-100326
Titolo
A Genetic Algorithm Based System to enable the Production of Transparent and Interpretable Fuzzy Logic Based Classifiers for the Modeling and Prediction of Complex Financial Data
Struttura
INGEGNERIA DELL'INFORMAZIONE
Corso di studi
INGEGNERIA INFORMATICA PER LA GESTIONE D'AZIENDA
Commissione
relatore Prof. Marcelloni, Francesco
correlatore Prof.ssa Lazzerini, Beatrice
Parole chiave
  • prediction
  • classifier
  • fuzzy logic
  • genetic algorithm
  • financial data
Data inizio appello
09/05/2013;
Consultabilità
parziale
Data di rilascio
09/05/2053
Riassunto analitico
The amount of information that companies have to manage at the same time to get the best value for their business is growing very fast through the use of IT systems. Usually is not obvious what to do with all this amount of information. For this reason there were been developed many theories and “intelligent” application that are able to give aggregate useful data to the end user.
In this work, we developed a Fuzzy Logic Classification Evolutionary System (FLCES) that is able to get a dataset as input and classify the inputs as “bad” customer or “good” customer, in order to help the management of the company to develop the appropriate strategy. We developed two different approaches that can be used individually or can be used in a cascade structure. The proposed FLCES approaches are based on a genetic algorithm that aims to increase the performance of the fuzzy classifier as well as to increase the interpretability of the rule base belonging to the fuzzy classifier. The implemented approach optimizes the length of the rules, making them more interpretable, and hence, improving the performance. A new approach for gene encoding has also been implemented reducing memory, speed and time consumption.
File