Machine Learning for Multimedia Movie Success Prediction

• Ano de defesa: 2023
• Autor(a) principal: Souza, Thais Luiza Donega e
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Análise de Sentimento; Filme; Image Processing; Movie; Multimodal; Processamento de Imagem; Sentiment Analysis
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/100/100131/tde-16012024-220439/

This study\'s primary goal is to predict, through computational techniques, the success of a film in its financial aspect from visual, textual, and metadata available before its release. The thesis is made of three main investigations. In the first article, machine learning algorithms are applied to update the literature using only metadata available to producers before a film release. In the second, a Convolutional Neural Network is built to extract information from movie posters; these results are analyzed from an economic perspective and might shed light on creating cheap and appealing marketing visual content. In the third, consumer sentiment analysis is performed to deal with textual data from the cast, drawing sentiments over actors to explore their impact on movie success. Although estimating a film\'s box office success is a widely explored topic in the literature, this is the first study to approach the problem in its multimodal form, as it indeed is. In addition, little is known about the advertising that surrounds a movie or its content; neither casting measures extracted directly from the audiences. Therefore, this work is the first to propose a decision support tool that answers the movie uncertainty problem with all the available data before production