Upgrade of the SFIA clustering module: weak signals for anticipative intelligence

• Grant number: 12/11917-0
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: September 01, 2012
• End date: August 31, 2013
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computer Systems
• Principal Investigator: Ana Estela Antunes da Silva
• Grantee: Thomas Wiliam Norio Hayashi
• Host Institution: Faculdade de Tecnologia (FT). Universidade Estadual de Campinas (UNICAMP). Limeira , SP, Brazil

Abstract

This project aims to upgrade the clustering module of the SFIA (Weak Signals for Anticipative Intelligence) tool which was the subject of a Fapesp project (process number: 2011/08696-9) which ends next July/2012. This is a tool to support the organizational decision-making process based on anticipative intelligence information. Anticipative intelligence is based on the extraction, storage, and analysis of information that anticipates events and, consequently, put organizations ahead of their rivals. A weak signal is anticipative information. It is a short sentence resulting from real observations. A clustering module has been developed in the SFIA tool based on the k-medoids clustering algorithm to group similar weak signals. The algorithm implements a similarity function whose elements are syntactic elements of weak signals. The function is based on the Jaccard similarity function. This project aims to upgrade the clustering module by taking into account two main aspects: the quality of weak signal groups and the representation power of the similarity function. In order to do so, this project proposes the implementation of an extension to the k-medoids algorithm. Some examples of extensions are bisecting k-medoids and constraint k-medoids. Regarding the similarity function aspect, the function used in the initial project is composed of common and distinct syntactic elements between pairs of weak signals. One of the evaluation criteria will be the pertinency analysis of such elements and new elements to the similarity function. The main objectives of this project can be summarized as: a study of the family of k-medoids algorithms; implementation of an extension of these algorithms; performance comparison of the two algorithms (the initial classic one and its extension), evaluation of the similarity function components. (AU)