| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 3.49 MB | Adobe PDF |
Advisor(s)
Abstract(s)
Ao longo dos últimos anos, as regras de associação têm assumido um papel relevante
na extracção de informação e de conhecimento em base de dados e vêm com isso auxiliar
o processo de tomada de decisão.
A maioria dos trabalhos de investigação desenvolvidos sobre regras de associação
têm por base o modelo de suporte e confiança. Este modelo permite obter regras de
associação que envolvem particularmente conjuntos de itens frequentes.
Contudo, nos últimos anos, tem-se explorado conjuntos de itens que surgem com
menor frequência, designados de regras de associação raras ou infrequentes. Muitas das
regras com base nestes itens têm particular interesse para o utilizador. Actualmente a
investigação sobre regras de associação procuram incidir na geração do maior número
possível de regras com interesse aglomerando itens raros e frequentes.
Assim, este estudo foca, inicialmente, uma pesquisa sobre os principais algoritmos
de data mining que abordam as regras de associação.
A finalidade deste trabalho é examinar as técnicas e algoritmos de extracção de
regras de associação já existentes, verificar as principais vantagens e desvantagens dos
algoritmos na extracção de regras de associação e, por fim, desenvolver um algoritmo
cujo objectivo é gerar regras de associação que envolvem itens raros e frequentes.
Over the past few years, association rules have taken an important paper in extracting information and knowledge from database, which helps the decision-making process. The most of the investigation works of in association rules is essentially based on the model of support and confidence. This model enables to extract association rules particularly related to frequent items. However, in recent years, the need to explore less frequent itemsets, called rare or unusual association rules, has increased. Many of these rules that involve infrequent items are considered to have particular interest for the user. Recently, efforts on the research of association rules have tried to generate the largest possible number of interest rules agglomerating rare and frequent items. This way, this study initially seals a research on the main algorithms of date mining that approach the association rules. An association rule is considered to be rare when it is formed by frequent and unusual items or unusual items only. The purpose of this study is to examine not only the techniques and algorithms for the extraction of association rules that already exist, but also the main advantages and disadvantages of the algorithms in the mining of association rules, and finally to develop an algorithm whose objective is to generate association rules that involve rare and frequent items.
Over the past few years, association rules have taken an important paper in extracting information and knowledge from database, which helps the decision-making process. The most of the investigation works of in association rules is essentially based on the model of support and confidence. This model enables to extract association rules particularly related to frequent items. However, in recent years, the need to explore less frequent itemsets, called rare or unusual association rules, has increased. Many of these rules that involve infrequent items are considered to have particular interest for the user. Recently, efforts on the research of association rules have tried to generate the largest possible number of interest rules agglomerating rare and frequent items. This way, this study initially seals a research on the main algorithms of date mining that approach the association rules. An association rule is considered to be rare when it is formed by frequent and unusual items or unusual items only. The purpose of this study is to examine not only the techniques and algorithms for the extraction of association rules that already exist, but also the main advantages and disadvantages of the algorithms in the mining of association rules, and finally to develop an algorithm whose objective is to generate association rules that involve rare and frequent items.
Description
Keywords
Regras de associação Itens frequentes Itens raros Association rules Frequent itemsets Rare itemsets
Citation
Publisher
Instituto Politécnico do Porto. Instituto Superior de Engenharia do Porto