The research activity aims to investigate the identification of suitable methods to support the decision-making process characterized by elements of complexity. In the real context, the activity of making decisions involves a certain degree of responsibility and a certain level of ever increasing risks due to the growing complexity of the systems. In the prefigured context, the role of the decision-maker becomes increasingly difficult to the point of seeking different methods and models, thus creating what we could define as a decision-making dashboard useful for making decisions in different contexts, to support and manage extremely complex decision-making processes. Following a careful analysis of the literature and of the various research currents conducted in the field of decision theory, the relevant theory of bounded rationality emerged, introduced by the Nobel Prize for Economics Herbert Simon, with its cyclical modeling of the decision-making process ; in contrast to the classical decision theory whose conceptual structure is mainly based on the Decision Maker's perspective of absolute rationality. It also emerges that the main element that characterizes the decision-making process is the level of information.The decision-making process is therefore more difficult when dealing with incomplete, imprecise and subjective information, or when the information is fuzzy. In this environment, fuzzy sets lend themselves well to characterizing the different types of information, making it possible to express ad hoc models in real contexts. In relation to these insights, the research work led to the presentation of a hybrid model between the fuzzy environment and artificial neural networks entitled "Hybrid Fuzzy Differential System and Artificial Neural Networks: Some Issues in Economics". The analysis originates from the fuzzy differential equations proposed to manage the uncertainty due to incomplete information, which appear in many decision analysis models allowing to better describe deterministic phenomena of the real world. The work proposes an approach to solve hybrid fuzzy differential equations based on feed-forward neural networks. This method shows that the use of neural networks provides solutions with good generalization and high precision. By comparing the results of this work with the results obtained with numerical methods, it emerged that the proposed model provides more accurate approximations. An interesting research on RSA public key cryptography is also proposed, useful in terms of information security which, as seen previously, represent an element of fundamental importance in the decision-making process, worthy of particular attention. Further elements have also been identified which determine a certain degree of complexity in the decision-making analysis, in particular, in the real context of decision making there are multiple scenarios, uncontrollable aspects therefore an uncertain environment; complex preferences, i.e. insufficient to define a concept of optimal, of best; multiple decision makers, with potentially conflicting preferences. In this perspective, the Multicriteria Decision Analysis has proved to be a valid methodology to help the decision maker. Relevant information emerged through bibliometric analysis of multi-criteria approaches: the editorial path has recorded an upward trend, the last few years with over 2500 publications; China represents the main country in terms of major contribution to research on multi-criteria methods; the method most used in the application field and which has recorded the greatest scientific production is the method called Analytic Hierarchy Process (AHP). The research activity focused on the latter data by introducing a new vision of the method. The method proposed in this work consists in the implementation of the AHP-Short multicriteria method and the analysis of predictive models of Machine Learning. From a research analysis, classification can be performed through logistic regression: it identifies predicts the probability of an observation belonging to the class. With this thesis work, attention is placed on the identification of tools to be included in a decision dashboard to give the best contribution to the mechanism characterized by elements of complexity
L’attività di ricerca si propone di indagare circa l’identificazione di idonei metodi di supporto al processo decisionale caratterizzato da elementi di complessità. Nel contesto reale l'attività di decidere comporta un certo grado di responsabilità e un determinato livello di rischi sempre maggiori dovuti alla crescente complessità dei sistemi. Nel contesto prefigurato il ruolo del decisore diventa sempre più difficile tanto da ricercare differenti metodi e modelli, creando quindi quello che potremmo definire un cruscotto decisionale utile ad assumere delle decisioni in diversi contesti, per sostenere e gestire processi decisionali estremamente complessi. A seguito di un’attenta analisi della letteratura e delle diverse correnti di ricerca condotte in ambito della teoria delle decisioni, è emersa la rilevante teoria della razionalità limitata introdotta dal premio Nobel per l’economia Herbert Simon, con la sua modellizzazione ciclica del processo decisionale; in contrapposizione alla teoria classica decisionale la cui struttura concettuale è fondata principalmente sulla prospettiva di razionalità assoluta del Decision Maker. Emerge inoltre che l’elemento principale che caratterizza il processo decisionale è il livello di informazione. Il processo decisionale è pertanto più difficile quando si ha a che fare con informazione incompleta, imprecisa e soggettiva, ovvero quando l’informazione è fuzzy. In tale ambiente gli insiemi fuzzy ben si prestano a caratterizzare i diversi tipi di informazione, permettendo di esplicitare modelli ad hoc a contesti reali. In relazione a tali approfondimenti il lavoro di ricerca ha condotto alla presentazione di un modello ibrido tra l’ambito fuzzy e le reti neurali artificiali dal titolo “Hybrid Fuzzy Differential System and Artificial Neural Networks: Some Issues in Economics”. L’analisi ha origine dalle equazioni differenziali fuzzy proposte per gestire l'incertezza dovuta a informazioni incomplete, che appaiono in molti modelli di analisi decisionale permettendo di meglio descrivere fenomeni deterministici del mondo reale. Il lavoro propone un approccio per risolvere le equazioni differenziali fuzzy ibride basate sulle reti neurali feed-forward. Questo metodo mostra che l’utilizzo delle reti neurali fornisce soluzioni con una buona generalizzazione e un’elevata precisione. Confrontando i risultati di questo lavoro rispetto ai risultati ottenuti con metodi numerici è emerso che il modello proposto fornisce approssimazioni più accurate. Viene inoltre proposta un'interessante ricerca sulla crittografia a chiave pubblica RSA, utile in termini di sicurezza delle informazioni che, come visto in precedenza rappresentano un elemento di fondamentale importanza nel processo decisionale, meritevole di particolare attenzione. Sono stati inoltre individuati ulteriori elementi che determinano un certo grado di complessità nell’analisi decisionale, in particolare, nel contesto reali di assunzione delle decisioni si registrano scenari molteplici, aspetti non controllabili dunque un ambiente incerto; preferenze complesse, cioè insufficienti a definire un concetto di ottimo, di migliore; decisori plurimi, con preferenze potenzialmente in conflitto tra loro. In tale prospettiva, la Multicriteria Decision Analysis è risultata essere una valida metodologia di aiuto al decision maker. Attraverso analisi bibliometrica degli approcci multicriteriali sono emerse informazioni rilevanti: il percorso editoriale ha registrato la tendenza al rialzo, gli ultimi anni con oltre 2500 pubblicazioni; la Cina rappresenta il principale paese in termini di maggior contributo alla ricerca sui metodi multicriteri; il metodo più utilizzato in ambito applicativo e che ha registrato una maggiore produzione scientifica è il metodo denominato Analytic Hierarchy Process (AHP). L’attività di ricerca si è soffermata su quest’ultimo dato introducendo una nuova visione del metodo. Il metodo proposto in questo lavoro consiste nell'implementazione del metodo multicriterio AHP-Short e l'analisi di modelli predittivi di Machine Learning. Da un’analisi di ricerca, la classificazione può essere effettuata attraverso la regressione logistica: identifica prevede la probabilità di appartenenza di un’osservazione alla classe. Con il presente lavoro di tesi si pone l’attenzioni sull’individuazione di strumenti da inserire in un cruscotto decisionale per dare il miglior contributo al meccanismo caratterizzato da elementi di complessità
The complexity of decisions mechanism design: decision support tools / Merenda, Domenica Stefania. - (2023 Apr 19).
The complexity of decisions mechanism design: decision support tools
Merenda, Domenica Stefania
2023-04-19
Abstract
The research activity aims to investigate the identification of suitable methods to support the decision-making process characterized by elements of complexity. In the real context, the activity of making decisions involves a certain degree of responsibility and a certain level of ever increasing risks due to the growing complexity of the systems. In the prefigured context, the role of the decision-maker becomes increasingly difficult to the point of seeking different methods and models, thus creating what we could define as a decision-making dashboard useful for making decisions in different contexts, to support and manage extremely complex decision-making processes. Following a careful analysis of the literature and of the various research currents conducted in the field of decision theory, the relevant theory of bounded rationality emerged, introduced by the Nobel Prize for Economics Herbert Simon, with its cyclical modeling of the decision-making process ; in contrast to the classical decision theory whose conceptual structure is mainly based on the Decision Maker's perspective of absolute rationality. It also emerges that the main element that characterizes the decision-making process is the level of information.The decision-making process is therefore more difficult when dealing with incomplete, imprecise and subjective information, or when the information is fuzzy. In this environment, fuzzy sets lend themselves well to characterizing the different types of information, making it possible to express ad hoc models in real contexts. In relation to these insights, the research work led to the presentation of a hybrid model between the fuzzy environment and artificial neural networks entitled "Hybrid Fuzzy Differential System and Artificial Neural Networks: Some Issues in Economics". The analysis originates from the fuzzy differential equations proposed to manage the uncertainty due to incomplete information, which appear in many decision analysis models allowing to better describe deterministic phenomena of the real world. The work proposes an approach to solve hybrid fuzzy differential equations based on feed-forward neural networks. This method shows that the use of neural networks provides solutions with good generalization and high precision. By comparing the results of this work with the results obtained with numerical methods, it emerged that the proposed model provides more accurate approximations. An interesting research on RSA public key cryptography is also proposed, useful in terms of information security which, as seen previously, represent an element of fundamental importance in the decision-making process, worthy of particular attention. Further elements have also been identified which determine a certain degree of complexity in the decision-making analysis, in particular, in the real context of decision making there are multiple scenarios, uncontrollable aspects therefore an uncertain environment; complex preferences, i.e. insufficient to define a concept of optimal, of best; multiple decision makers, with potentially conflicting preferences. In this perspective, the Multicriteria Decision Analysis has proved to be a valid methodology to help the decision maker. Relevant information emerged through bibliometric analysis of multi-criteria approaches: the editorial path has recorded an upward trend, the last few years with over 2500 publications; China represents the main country in terms of major contribution to research on multi-criteria methods; the method most used in the application field and which has recorded the greatest scientific production is the method called Analytic Hierarchy Process (AHP). The research activity focused on the latter data by introducing a new vision of the method. The method proposed in this work consists in the implementation of the AHP-Short multicriteria method and the analysis of predictive models of Machine Learning. From a research analysis, classification can be performed through logistic regression: it identifies predicts the probability of an observation belonging to the class. With this thesis work, attention is placed on the identification of tools to be included in a decision dashboard to give the best contribution to the mechanism characterized by elements of complexityFile | Dimensione | Formato | |
---|---|---|---|
Merenda Domenica Stefania.pdf
accesso aperto
Tipologia:
Tesi di dottorato
Licenza:
DRM non definito
Dimensione
20.85 MB
Formato
Adobe PDF
|
20.85 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.