Banca de QUALIFICAÇÃO: CASSIA DE SOUZA CARVALHO

Uma banca de QUALIFICAÇÃO de DOUTORADO foi cadastrada pelo programa.
DISCENTE : CASSIA DE SOUZA CARVALHO
DATA : 05/06/2019
HORA: 14:00
LOCAL: sala 303, 3º andar do Bloco B, campus Santo André
TÍTULO:

An Interpretable Hybrid Recommender System

PÁGINAS: 60
GRANDE ÁREA: Ciências Exatas e da Terra
ÁREA: Ciência da Computação
RESUMO:

Hybrid Recommender Systems deliver recommendations by combining the collective information of users' preference, such as ratings, and side information, such as user and item features. An interpretable model using those features can increase the value of recommendations, allowing explanations about the reasons a user should pursue an item, to understand the overall consumption behavior of users or to plan marketing campaigns. However, transforming categorical features into numerical often increases the data dimensionality, and consequently, raises the computational cost. In this project, we present a new recommender system framework, that consists of different processing steps with the main characteristic of preserving the interpretability of the final model. The basis of this framework is the CARFRE categorical transformation technique capable of keeping the dimensionality of the data set without losing the meaning of the original feature space. This technique will be coupled with a Boolean Function Synthesis algorithm in order to find interesting combinations of categorical feature, a data clustering step, to allow the creation of more focused models, the use of linear models on the recommendation end of the system and, finally, a translator of model to natural language to ensure a proper explanation of the recommendation or generating reports of the consumer behaviors.

MEMBROS DA BANCA:
Presidente - 1932365 - FABRICIO OLIVETTI DE FRANCA
Interno - 1673092 - RONALDO CRISTIANO PRATI
Interno - 2376122 - THIAGO FERREIRA COVOES