Multi-Relational Model Tree Induction Tightly-Coupled with a Relational Database

Abstract

Multi-Relational Data Mining (MRDM) refers to the process of discovering implicit, previously unknown and potentially useful information from data scattered in multiple tables of a relational database. Following the mainstream of MRDM research, we tackle the regression where the goal is to examine samples of past experience with known continuous answers (response) and generalize future cases through an inductive process. Mr-SMOTI, the solution we propose, resorts to the structural approach in order to recursively partition data stored into a tightly-coupled database and build a multi-relational model tree which captures the linear dependence between the response variable and one or more explanatory variables. The model tree is top-down induced by choosing, at each step, either to partition the training space or to introduce a regression variable in the linear models with the leaves. The tight-coupling with the database makes the knowledge on data structures (foreign keys) available free of charge to guide the search in the multi-relational pattern space. Experiments on artificial and real databases demonstrate that in general Mr-SMOTI outperforms both SMOTI and M5' which are two propositional model tree induction systems, and TILDE-RT which is a state-of-art structural model tree induction system. A. Appice et al. / Multi-Relational Model Tree Induction Tightly-Coupled with a Relational DatabaseSome examples of applications are found in mining the web, which can be seen as a massive graph of linked pages [19], in social network analysis, which focuses on modeling the relationships between social (individuals, groups, organizations) and computational (computers and web sites) actors [38], in computational biology, which studies interactions between components (genes and proteins) of the living cells [33], and in mining geo-referenced objects, which are characterized by implicit spatial relationships defined by geometry and positioning [28].Multi-relational data mining (MRDM) is a multi-disciplinary field which provides sound methods for the discovery of relational patterns that involve multiple relations and are stated in a more expressive language (e.g., predicate calculus and SQL) than patterns defined on a single data table.Many data mining tasks (e.g., classification, clustering, association analysis) have been adapted to a multi-relational setting. In particular, several studies have been reported for the regression task, which is the main focus of this paper. In the regression task, some labeled observations (training cases) are given, which are instances of an unknown continuous function f , and the problem is to build a general model that approximates the function f and can be subsequently used to make predictions on new unlabeled observations.Handling multi-relational data adds difficulties to the regression task. First of all, data available in distinct database relations describe different objects (or entities) involved in the phenomenon under study. These objects play different roles, and it is neces...