|Tipo di tesi||Tesi di laurea magistrale|
|Titolo||Hidden Markov Models per la predizione delle transizioni di stato di pazienti SARS-CoV-2|
|Titolo in inglese||Hidden Markov Models to predict the transitions of SARS-CoV-2 patients' state|
|Struttura||Dipartimento di Scienze Fisiche, Informatiche e Matematiche|
|Corso di studi||INFORMATICA (D.M. 270/04)|
|Data inizio appello||2020-10-21|
|Disponibilità||Accessibile via web (tutti i file della tesi sono accessibili)|
Lo stato di emergenza sollevato dall'epidemia SARS–CoV–2 del 2019 ha generato il bisogno di modelli teorici ed applicati, capaci di trattare con l'ignoto e la mancanza di conoscenza relativi a catastrofi così improvvise.
The state of emergency raised by the SARS–CoV–2 pandemic in 2019 generated the need of theoretical and applied models, able to deal with the unknown and the lack of knowledge related to such a sudden catastrophe. From the medical side, several hypotheses are usually available; in any case, these are rarely sufficient to understand, model and predict the whole complex dynamic happening inside an ill patient. Given the availability of data collected on a (fortunately) small population of patients, several machine learning approaches are not applicable, for two main reasons: • a small amount of data does not allow to generalize and thus learn well like a huge one actually does; • doctors prefer to trust models, which besides giving a good result also explain it; this way the result can be also checked from a medical point of view. Hidden Markov Models can be an answer to the above issues because: given several important (accordingly to hypotheses made by doctors) observations, they can build chains of hidden unobservable states which actually model the observed phenomena. Moreover, since they are by design built on top of medical advices, they do not need to provide a further explanation or proof. In addition to these points of strength, Hidden Markov Models better allow, compared to other machine learning models, to deal with missing or incomplete data, like the one often collected in a hurry during pandemics, where the absolute and unquestionable priority is saving human beings.