Multi relational-upgraded methods: Classification and analysis in supervised learning

Abstract

Today, Relational databases are used to store structured and complex data. They consist of multiple relations that are linked together conceptually via entity-relationship links. While the need to analyze these complex and structured data now have been increased, but many traditional learning techniques mine a single table as input and could not meet this requirement. So Multi-relational data mining methods are proposed which they search for the useful patterns involved in multiple relations in a relational database. Classification is one of the most popular tasks in data mining, so numerous studies have been done on solving multi-relational classification problems. Due to variety and plenty of proposed methods in this field, it seems useful to understand better the classification of these methods and the differences between them. In this work, we provide an analysis focuses on Relational-upgraded methods that upgrade traditional supervised classification algorithms such as KNN, Naive Bayesian, decision tree, genetic algorithms into the multi relational classification. We classify these methods according to traditional methods that are upgraded, then review and analyze the existing multi relational classification techniques in each group. Therefore, this study presents a formal classification and analysis that provides a useful roadmap for new researchers in the area of multi relational classification.