Co-clustering in ensembles of collaborative filters

Abstract

The advances in information technology are promoting the generation and storage of ever-increasing amount of data, produced by virtually all fields of human activity. So, there are distinguished demands for the analysis and knowledge extraction of these data, aiming at better understanding the processes involved, detecting trends and supporting strategic plans for businesses and corporations. Recommender systems have been used for this purpose, and have proven capable of promoting productivity increase, maximization of customer satisfaction and saving of resources. The main objective of this research project is the implementation of scalable computational tools for the synthesis of ensembles of collaborative filters, using: (1) random projections, for dimension reduction in datasets; (2) co-clustering, to provide multiple perspectives for data analysis in collaborative filtering; (3) machine learning techniques, such as k-nearest neighbors, for the synthesis of ensembles of collaborative filters. It will be taken as case studies: (1) Imputation for missing data; and (2) Support systems for pattern detection, such as anomalies, in brain activity data, as well as the design of customized human-machine interfaces. Both applications deal with areas of high scientific and corporate interests, and the case study (2) is linked to the project entitled CEPID BRAINN (Brazilian Research Institute for Neuroscience and Neurotechnology), which makes the proposal of high potential impact on society.