Files in this item

FilesDescriptionFormat

application/pdf

application/pdfGHAYOORI-DISSERTATION-2020.pdf (6MB)Restricted to U of Illinois
(no description provided)PDF

Description

Title:Budget allocation and optimal use of resources in four different contexts: Data centers, viral marketing, recommendation systems, and the fight with the HIV epidemic
Author(s):Ghayoori, Arash
Director of Research:Nagi, Rakesh
Doctoral Committee Chair(s):Nagi, Rakesh
Doctoral Committee Member(s):Beck, Carolyn; Etesami, S. Rasoul; Jacobson, Sheldon
Department / Program:Industrial&Enterprise Sys Eng
Discipline:Industrial Engineering
Degree Granting Institution:University of Illinois at Urbana-Champaign
Degree:Ph.D.
Genre:Dissertation
Subject(s):Social networks
Viral marketing
Influence maximization
Random graph models
Diffusion models
Recommender Systems
Interaction Analysis
Trust Analysis
Product Popularity
Abstract:Optimal allocation of the available budget and resources in a manner that maximizes a pre-determined utility function over a certain population is of significant importance in various fields such as marketing, global health, social and political sciences, E-commerce, etc. In this thesis, we attempt to solve this problem, under clearly specified assumptions, over four contexts. We first attempt at tackling this problem in the context of data centers. Given that adding or removing a network switch can be very costly, we developed an algorithm with high throughput that allows us to add/remove switches very easily. Secondly, we attempt at tackling this problem in the field of viral marketing. We found a pessimistic bound on the return on investment for a potential investor. Specifically, we found a lower bound on the probability that a certain percentage of the underlying network would be dominated by a target time. This would show the investor the worst possible scenario for his/her investment in order to reach his target, which facilitates informed decision making. We also designed a seed selection algorithm, under the linear threshold (LT) diffusion model, and the exact minimum required number of nodes that need to be activated in the underlying social network for ultimate market domination of a product. In the third part of our work, given the data that was available to us from Rwanda on their planned efforts to fight HIV spread, by targeting the most disturbed sub-population, female sex workers (FSWs), using the existing prevention/treatment methods such as promotion of condom use, safe-sex practice, and PrEP to the HIV$-$ portion, and providing ART to the HIV$+$ portion of the population. Using the data, we showed the optimal allocation of resources to each one of these methods, so that the overall health benefits were maximized. We showed this using extensive simulation studies. Lastly, in the final part of our work, we designed a recommendation system, that rather than assuming independence for all users, considers the underlying social influence from people ''close'' to the user, as part of the decision making process in what to recommend.
Issue Date:2020-09-10
Type:Thesis
URI:http://hdl.handle.net/2142/109472
Rights Information:Copyright 2020 Arash Ghayoori
Date Available in IDEALS:2021-03-05
Date Deposited:2020-12


This item appears in the following Collection(s)

Item Statistics