Social networks are used today for various purposes, such as communication, information, and advertising, but they present various problems that are currently being studied. This thesis addresses one of the problems inherent to advertising in social networks: the optimal allocation of ad, short for advertising, by an agent on the network's user nodes. To this end, we first define the model of the system, which turns out to be composed of two main elements: the agent that must allocate the ad and the social network on which the allocation is to be made. Subsequently, the characteristics of each of the two components of the model are identified and, in the social network, the hierarchical levels of which it is composed and between which relations are established are identified through the use of graph theory: the users and the posts, i.e. the content published in the social network by the users themselves. The characteristics of the latter are also defined, which, together with those of the agent, are used for the formulation of the optimisation problem, whose objective function requires the estimation, for each user under analysis, of the expected views of the post containing the ad. This is produced using the methods made available by time series analysis, in particular the ARIMA (AutoRegressive Integrated Moving Average) prediction model. The consequent formulation and resolution of the optimisation problem, leads to the identification of the combination of user nodes on which the allocation of the ad is optimal for the agent, considering the budget it has available and the cost of the allocation itself, which is different for each user. Finally, the proposed model is tested on real data, through a simulator implemented in a Python environment and based on the use of the pmdarima and CVXPY libraries.
I social network sono oggi utilizzati per vari scopi, quali comunicazione, informazione, pubblicizzazione, ma presentano diverse problematiche attualmente oggetto di studio. Questa tesi affronta uno dei problemi inerenti la pubblicità nei social network: l’allocazione ottimale di ad, abbreviazione di advertising, da parte di un agente sui nodi utente della rete. A tal fine, si definisce innanzitutto il modello del sistema, che risulta essere composto di due elementi principali: l’agente che deve allocare l’ad e il social network su cui effettuare l’allocazione. Successivamente, si identificano le caratteristiche di ognuna delle due componenti del modello e, nel social network, si individuano, tramite l’utilizzo della teoria dei grafi, i livelli gerarchici di cui questo è composto e tra cui si instaurano delle relazioni: gli utenti e i post, ovvero i contenuti pubblicati nel social network dagli utenti stessi. Anche di questi ultimi si definiscono le caratteristiche che, unitamente a quelle dell’agente, sono utilizzate per la formulazione del problema di ottimizzazione, nella cui funzione obiettivo è richiesta la stima, per ogni utente oggetto di analisi, delle visualizzazioni attese del post contenente l’ad. Questa è prodotta utilizzando i metodi messi a disposizione dall’analisi delle serie temporali, in particolare il modello di previsione ARIMA (AutoRegressive Integrated Moving Average). La conseguente formulazione e risoluzione del problema di ottimizzazione, porta quindi all’individuazione della combinazione di nodi utente sui cui l’allocazione di ad risulta essere ottima per l’agente, considerando il budget che questo ha a disposizione e il costo dell’allocazione stessa, diverso per ciascun utente. Infine, il modello proposto è testato su dati reali, tramite un simulatore implementato in ambiente Python e basato sull’utilizzo delle librerie pmdarima e CVXPY.
ALLOCAZIONE OTTIMALE DI RISORSE PER PUBBLICITÀ ONLINE SU RETI SOCIALI
MARSEGLIA, GABRIELLA
2022/2023
Abstract
Social networks are used today for various purposes, such as communication, information, and advertising, but they present various problems that are currently being studied. This thesis addresses one of the problems inherent to advertising in social networks: the optimal allocation of ad, short for advertising, by an agent on the network's user nodes. To this end, we first define the model of the system, which turns out to be composed of two main elements: the agent that must allocate the ad and the social network on which the allocation is to be made. Subsequently, the characteristics of each of the two components of the model are identified and, in the social network, the hierarchical levels of which it is composed and between which relations are established are identified through the use of graph theory: the users and the posts, i.e. the content published in the social network by the users themselves. The characteristics of the latter are also defined, which, together with those of the agent, are used for the formulation of the optimisation problem, whose objective function requires the estimation, for each user under analysis, of the expected views of the post containing the ad. This is produced using the methods made available by time series analysis, in particular the ARIMA (AutoRegressive Integrated Moving Average) prediction model. The consequent formulation and resolution of the optimisation problem, leads to the identification of the combination of user nodes on which the allocation of the ad is optimal for the agent, considering the budget it has available and the cost of the allocation itself, which is different for each user. Finally, the proposed model is tested on real data, through a simulator implemented in a Python environment and based on the use of the pmdarima and CVXPY libraries.| File | Dimensione | Formato | |
|---|---|---|---|
|
MARSEGLIA_tesi.pdf
File consultabile da postazioni biblioteca
Descrizione: Tesi di laurea triennale in ingegneria gestionale di Marseglia Gabriella
Dimensione
1.65 MB
Formato
Adobe PDF
|
1.65 MB | Adobe PDF |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14253/64293