Implementation and evaluation of recommendation systems based on matrix factorization techniques

• Grant number: 16/25075-1
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: March 01, 2017
• End date: November 30, 2017
• Field of knowledge: Physical Sciences and Mathematics - Computer Science
• Principal Investigator: Solange Oliveira Rezende
• Grantee: Vítor Rodrigues Tonon
• Host Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil

Abstract

Users face difficulties in choosing products and services on the Web due to the wide variety of options available. Recommendation systems are systems that aim to help users identify items of interest in a set of options. Traditional approaches to recomender systems focus on recommending more relevant items to individual users, not taking into account their context. Among these techniques are the collaborative filtering techniques that use matrix factorization. Such techniques have become popular because they combine good scalability with good accuracy, and offer the flexibility to model various real-world situations. However, in many applications, it is also important to consider contextual information to make the recommendations. Studies show that the use of context improves the accuracy of the recommendation. Thus, methods based on matrix factorization that use contextual information are methods of recommendation systems that can present good accuracy. Therefore, in this project of Scientific Initiation, two approaches will be implemented of recommendation systems based on matrix factorization: (I) classical approach, which does not use context information and (II) a context-based approach. The two implementations will be made in the R language and embedded in the framework of recommendation systems, developed by Professor Dr. Marcos Aurélio Domingues. These implementations will be evaluated and made available. (AU)