On January 30, 2020, the World Health Organization (WHO) declared a state of international public health emergency due to the Coronavirus epidemic in China. On 11 March 2020, the director general of the WHO defined the spread of Covid-19 as a pandemic throughout the planet. Scientists from all over the world immediately began studying the virus for a vaccine, and more than 150 vaccines were tested. At the end of December, the vaccination campaign officially began in Europe, initially intended for health workers and frail people, and then progressively extended to the entire population over 5 years of age. However, since vaccinations were not deemed mandatory in Italy, it was necessary to identify those that were resistant against getting vaccinated due to various reasons. Thus, the creation of Personas to identify the main characteristics of people willing to be vaccinated was deemed of interest. The goal of this thesis is to compare multiple clustering and supervised machine learning methods in order to create Personas that stand out for their willingness, or lack thereof, to vaccinate against Covid-19. A survey, created by the University of Milan to investigate people's psychophysical conditions and their willingness to get vaccinated, was administered online, directed to the Italian population, in the months of January and February 2021. The data obtained from the survey were analysed through agglomerative and hierarchical clustering techniques. Further analyses were performed using supervised machine learning methods, creating classification models. Through statistical analysis, the differences within the datasets obtained by the previous methods were investigated, dividing them into significantly different clusters. Three Persona tables were thus obtained, which were compared with each other to evaluate their effectiveness in highlighting common characteristics among individuals who would not have been vaccinated against Covid-19. This thesis shows that supervised machine learning methods are comparable, in results, to statistical analysis methods and clustering techniques in identifying the main characteristics of Personas if a target variable is available. The goal of this thesis is to compare multiple clustering and unsupervised machine learning methods in order to create Personas that stand out for their willingness to vaccinate against Covid-19. A survey, created by the University of Milan to investigate people's psychophysical conditions and their willingness to get vaccinated, was administered online, directed to the Italian population, in the months of January and February 2021. The data obtained from the survey were analyzed through agglomerative and hierarchical clustering techniques. Further analyzes were performed using unsupervised machine learning methods, creating classification models. Through statistical analysis, the differences within the datasets obtained by the previous methods were investigated, dividing them into significantly different clusters. Three clusterizations were thus obtained, which were compared with each other to evaluate their effectiveness in highlighting common characteristics among individuals who would not have been vaccinated against Covid-19.
Il 30 Gennaio 2020 l’Organizzazione Mondiale della Sanità (OMS) ha dichiarato lo stato di emergenza internazionale di salute pubblica dovuta all’epidemia di Coronavirus in Cina. L’11 Marzo 2020 il direttore generale dell’OMS ha definito la diffusione del Covid-19 una pandemia in tutto il pianeta. A fine dicembre in Europa è iniziata ufficialmente la campagna vaccinale, inizialmente destinata ad operatori sanitari e persone fragili, per poi estendersi progressivamente a tutta la popolazione sopra i 5 anni. Tuttavia, poiché in Italia il vaccino non è stato reso obbligatorio, si è cercato di identificare i motivi che hanno spinto le persone a decidere di non vaccinarsi. Pertanto la creazione di Personas è stata considerata di interesse per identificare le caratteristiche principali delle persone disposte a vaccinarsi. L'obiettivo di questa tesi è quello di confrontare più metodi di clustering e machine learning supervisionato al fine di creare Personas che si distinguano per la loro volontà di vaccinarsi contro il Covid-19. Un questionario, appositamente creato dall’Università di Milano per indagare sulle condizioni psicofisiche delle persone e sulla loro volontà di vaccinarsi, è stato somministrato online, diretto alla popolazione italiana, nei mesi di Gennaio e Febbraio 2021. I dati ottenuti dal questionario sono stati analizzati tramite tecniche di clustering agglomerativo e gerarchico. Ulteriori analisi sono state eseguite tramite metodi di machine learning supervisionato, creando modelli di classificazione. Attraverso l’analisi statistica si sono andate ad approfondire le differenze all’interno dei dataset ottenuti dai precedenti metodi, suddividendoli in clusters significativamente differenti tra loro. Si sono così ottenute tre clusterizzazioni, che sono state confrontate tra loro per valutarne l’efficacia nell’evidenziare caratteristiche comuni tra gli individui che non si sarebbero vaccinati. Questa tesi mostra come i metodi di machine learning supervisionato sono comparabili, nei risultati, alle tecniche di clustering e analisi statistica nell’identificare le caratteristiche principali delle Personas, se è presente una variabile target. L'obiettivo di questa tesi è quello di confrontare più metodi di clustering e unsupervised machine learning al fine di creare Personas che si distinguano per la loro volontà di vaccinarsi contro il Covid-19. Un questionario, appositamente creato dall’Università di Milano per indagare sulle condizioni psicofisiche delle persone e sulla loro volontà di vaccinarsi, è stato somministrato online, diretto alla popolazione italiana, nei mesi di Gennaio e Febbraio 2021. I dati ottenuti dal questionario sono stati analizzati tramite tecniche di clustering agglomerativo e gerarchico. Ulteriori analisi sono state eseguite tramite metodi di machine learning non supervisionato, creando modelli di classificazione. Attraverso l’analisi statistica si sono andate ad approfondire le differenze all’interno dei dataset ottenuti dai precedenti metodi, suddividendoli in clusters significativamente differenti tra loro. Si sono così ottenute tre clusterizzazioni, che sono state confrontate tra loro per valutarne l’efficacia nell’evidenziare caratteristiche comuni tra gli individui che non si sarebbero vaccinati contro il Covid-19.
Evaluation of statistical analysis methods and machine learning techniques to develop Personas for willingness of vaccination against Covid-19
ONOFRI, FRANCESCA
2021/2022
Abstract
On January 30, 2020, the World Health Organization (WHO) declared a state of international public health emergency due to the Coronavirus epidemic in China. On 11 March 2020, the director general of the WHO defined the spread of Covid-19 as a pandemic throughout the planet. Scientists from all over the world immediately began studying the virus for a vaccine, and more than 150 vaccines were tested. At the end of December, the vaccination campaign officially began in Europe, initially intended for health workers and frail people, and then progressively extended to the entire population over 5 years of age. However, since vaccinations were not deemed mandatory in Italy, it was necessary to identify those that were resistant against getting vaccinated due to various reasons. Thus, the creation of Personas to identify the main characteristics of people willing to be vaccinated was deemed of interest. The goal of this thesis is to compare multiple clustering and supervised machine learning methods in order to create Personas that stand out for their willingness, or lack thereof, to vaccinate against Covid-19. A survey, created by the University of Milan to investigate people's psychophysical conditions and their willingness to get vaccinated, was administered online, directed to the Italian population, in the months of January and February 2021. The data obtained from the survey were analysed through agglomerative and hierarchical clustering techniques. Further analyses were performed using supervised machine learning methods, creating classification models. Through statistical analysis, the differences within the datasets obtained by the previous methods were investigated, dividing them into significantly different clusters. Three Persona tables were thus obtained, which were compared with each other to evaluate their effectiveness in highlighting common characteristics among individuals who would not have been vaccinated against Covid-19. This thesis shows that supervised machine learning methods are comparable, in results, to statistical analysis methods and clustering techniques in identifying the main characteristics of Personas if a target variable is available. The goal of this thesis is to compare multiple clustering and unsupervised machine learning methods in order to create Personas that stand out for their willingness to vaccinate against Covid-19. A survey, created by the University of Milan to investigate people's psychophysical conditions and their willingness to get vaccinated, was administered online, directed to the Italian population, in the months of January and February 2021. The data obtained from the survey were analyzed through agglomerative and hierarchical clustering techniques. Further analyzes were performed using unsupervised machine learning methods, creating classification models. Through statistical analysis, the differences within the datasets obtained by the previous methods were investigated, dividing them into significantly different clusters. Three clusterizations were thus obtained, which were compared with each other to evaluate their effectiveness in highlighting common characteristics among individuals who would not have been vaccinated against Covid-19.| File | Dimensione | Formato | |
|---|---|---|---|
|
2022_07_Onofri.pdf
accessibile in internet per tutti
Dimensione
2.29 MB
Formato
Adobe PDF
|
2.29 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/190431