Um novo algoritmo imunológico artificial para agrupamento de dados

• Ano de defesa: 2010
• Autor(a) principal: Borges, Ederson
• Orientador(a): Silva, Leandro Nunes de Castro
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Presbiteriana Mackenzie
• Palavras-chave em Português: agrupamento de dados; diversidade; k-médias; rede imunológica artificial; sistemas imunológicos artificiais
• Palavras-chave em Inglês: clustering; diversity; k-means; artificial immune network; artificial immune systems
• Área do conhecimento CNPq: CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA
• Link de acesso: http://dspace.mackenzie.br/handle/10899/24414

Clustering is an important data mining task from the field of Knowledge Discovery in Databases. There are several algorithms capable of performing clustering tasks, and the most popular ones involve the calculation of a similarity or distance measure among objects from the database. Many algorithms can perform clustering in a simple and efficient manner, but have drawbacks as a way to get the optimal number of partitions and the possibility of getting stuck in local optima solutions. To try and reduce these drawbacks this dissertation proposes a new clustering algorithm based on Artificial Immune Systems. This algorithm is characterized by the generation of multiple simultaneous high quality solutions in terms of the number of partitions (clusters) for the database and the use of a cost function that explicitly evaluates the quality of partitions, minimizing the inconvenience of getting stuck in local optima. The algorithm was tested using four databases known in the literature and obtained satisfactory results in terms of the diversity of solutions, but has a high computational cost compared to other algorithms tested.