Each day more than millions of posts are published on social media networks and their active users are increasing exceedingly. These networks are massive sources of user-generated data, which can be used to extract valuable information in many fields. Some of these networks grant access to their data through their APIs, which facilitates many researchers to mine their data for the sake of analysis. In this thesis we tried to come up with innovative analysis approaches in dealing with cross-related events based on the aggregated data from social media networks. We studied a program that is a series of event, which took place in Milan during Expo 2015. The program is a collection of different events that feature both diversification points and mutual characteristics as well. The program is clustered into 4 groups of events (Performance, Art, Media, Science) and its dataset was gathered by probing into Instagram and Twitter APIs. The metrics that we used for assessing the data are based on the previous analysis techniques that have been applied to data derived from social media data. Our endeavor in this thesis was focused on performing the analysis in different stages of the program to understand the circumstances of each event and their connections with each other. The types of analysis that were applied include text mining, clustering, topic analysis, engagement analysis and correlation analysis. We extracted the temporal stream of out dataset, top trends of the events, features indicating user commitment and correlation between events based on their similarities. After that we employed a flexible and scalable visualization approach for modeling the outcomes. The results of the process of social media analytics gave us great insights on how each element of such program was carried, like events and their specifics and connections, and also how did the elements behave with respect to one another.
Ogni giorno milioni di messaggi vengono pubblicati sui social network e gli utenti attivi continuano ad aumentare esponenzialmente. Queste reti sono enormi fonti di dati generati dagli utenti, che possono essere utilizzati per estrarre informazioni preziose in molti campi di ricerca. Alcune di queste reti consentono l'accesso al dato attraverso le rispettive API, che facilitano a molti ricercatori il compito di estrarre i dati a scopo di analisi. In questa tesi si è cercato di individuare approcci di analisi innovativi per il trattamento di eventi cross-basate su dati aggregati provenienti dai social network. Abbiamo studiato un programma, rappresentato da una serie di eventi, che ha avuto luogo a Milano in occasione dell'Expo 2015. Il programma è una raccolta di diversi eventi che caratterizzano entrambi i punti di diversificazione e le caratteristiche comuni di investimento. Il programma è di tipo cluster ed organizzato in 4 gruppi di eventi (Performance, Arte, Media, Scienza) ed i dati sono stati raccolti utilizzando le API di Instagram e Twitter. Le metriche che abbiamo utilizzato per valutare i dati si basano sulle tecniche di analisi precedenti, applicate a dati derivati dai social network. Il nostro sforzo in questa tesi si è concentrato sull'analisi in diverse fasi del programma per comprendere le circostanze di ciascun evento e le loro interconnessioni. I tipi di analisi che sono state applicate comprendono l'estrazione di testo, il clustering, l'analisi dell'argomento, l'analisi dell'impegno e di correlazione. Abbiamo estratto il flusso temporale del set di dati, il top trend degli eventi, le caratteristiche che indicano l'impegno dell'utente e la correlazione tra gli eventi in base alle loro somiglianze. Dopo di che abbiamo adottato un approccio di visualizzazione flessibile e scalabile per la modellazione dei risultati. I risultati del processo di analisi dei social network ci ha fornito ottime intuizioni su ogni elemento di tale programma, rispettivamente ai singoli eventi, le loro specificità e le rispettive connessioni.
User engagement analysis of live events based on social media monitoring
JAFARI, ROKSANA
2015/2016
Abstract
Each day more than millions of posts are published on social media networks and their active users are increasing exceedingly. These networks are massive sources of user-generated data, which can be used to extract valuable information in many fields. Some of these networks grant access to their data through their APIs, which facilitates many researchers to mine their data for the sake of analysis. In this thesis we tried to come up with innovative analysis approaches in dealing with cross-related events based on the aggregated data from social media networks. We studied a program that is a series of event, which took place in Milan during Expo 2015. The program is a collection of different events that feature both diversification points and mutual characteristics as well. The program is clustered into 4 groups of events (Performance, Art, Media, Science) and its dataset was gathered by probing into Instagram and Twitter APIs. The metrics that we used for assessing the data are based on the previous analysis techniques that have been applied to data derived from social media data. Our endeavor in this thesis was focused on performing the analysis in different stages of the program to understand the circumstances of each event and their connections with each other. The types of analysis that were applied include text mining, clustering, topic analysis, engagement analysis and correlation analysis. We extracted the temporal stream of out dataset, top trends of the events, features indicating user commitment and correlation between events based on their similarities. After that we employed a flexible and scalable visualization approach for modeling the outcomes. The results of the process of social media analytics gave us great insights on how each element of such program was carried, like events and their specifics and connections, and also how did the elements behave with respect to one another.| File | Dimensione | Formato | |
|---|---|---|---|
|
final.pdf
accessibile in internet per tutti
Dimensione
15.48 MB
Formato
Adobe PDF
|
15.48 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/10589/119225