MURENA

A MUlti RElational aNAlyzer

LACAM @ Dipartimento di Informatica - Università degli Studi di Bari - Via Orabona, 4 -70126 Bari

A short description
The architecture
The distribution package
Project team
Related publications
FAQs

A short description

MURENA has been designed in order to efficiently support the user in the process of extracting knowledge from a relational database. In particular, the system supports relational data mining tasks from data stored in an ORACLE relational database.
Distinguishing features:
Relational Data Mining functionalities. MURENA embeds the Multi-relational Data mining systems Mr-SBC and Mr-SMOTI for classification and regression tasks respectively.
Distributed architecture. MURENA has been designed in order to allow multiple users to exploit its functionalities. MURENA interfaces every ORACLE local/remote database.
Portability. MURENA has been developed in java and uses the Enterprise Java Beans(TM) technology.
Space complexity. MURENA does not store the entire database in main memory, but directly queries the database when necessary.

Top of this page

The architecture

MURENA is a three tier client-server application that consists of three main components:

The data base level that interfaces an Oracle DBMS

The server, that is in charge of executing Data Mining algorithms and of the application of induced models.

The client, that is in charge of the user interfacing. It provides a graphical representation of the database schema and allows the user to execute Data Mining tasks.

The Server component can manage several clients simultaneously. In order to interface remote ORACLE databases by means of a JDBC connection, it is sufficient to specify the IP address, the port and the database SID.

Top of this page

MURENA Client-screenshot

By means of the user interface, the user can:

Database connection During this step, the user can open a connection to an ORACLE database. The user must specify: the database URL, the port number, the Database SID, the userid and the password. Once the connection is opened, the system shows the set of database schemas. The user can select the schema of interest and the system shows a graphical representation of the selected schema.

Data selection Scope of this step is the selection of the partition of the selected database schema to be used in the Data Mining process. In fact, MURENA uses all available tables and attributes (if the user does not explicitly exclude some).

Data Mining step This is the main task of MURENA. MURENA builds a prediction model from data selected in the previous steps. At now, in MURENA, two Multi relational Data Mining algorithms are implemented:

Mr-SBC builds a classification model (for discrete attributes prediction) (See related publications)

Mr-SMOTI builds a regression model (for continuous attributes prediction) (See related publications)

In both cases, the user has to specify the target attribute (the attribute to be predicted). Output of this task is a prediction model in XML format.

Model Application In this step the user can apply the induced model in order to predict the target attribute, given a database schema. If the target attribute is not unknown, the predictive accuracy of the induced model is evaluated.

Top of this page

The distribution package

MURENA Server is provided as a .jar file and can be executed on every machine with a JVM (1.4.0 or higher) installed.
Due to the use of the Application server, the client part is not available for download (if necessary, contact the contact person).

Read the (readme.txt file for the installation

Download the distribution package (murena.tar.gz, 1.3 MB).

A sample dataset is available for evaluation. Please, email the contact person for instructions.

Warning: The system MURENA is free for evaluation, research and teaching purposes, but not for commercial purposes.
Please Acknowledge

Top of this page

Project team

Project Leader

Donato Malerba

LACAM Staff

Annalisa Appice, Michelangelo Ceci

Students currently involved in the project

Nicola Barile

Previous collaborators

Vincenzo Colonna, Domenico Sacchi

Top of this page

Related publications

(in inverse chronological order)

M. Ceci, A. Appice, & D. Malerba (2004). Spatial Associative Classification at Different Levels of Granularity: A Probabilistic Approach, in J.-F. Boulicaut, F. Esposito, F. Giannotti, & D. Pedreschi (Eds.), Knowledge Discovery in Databases: PKDD 2004, Lecture Notes in Artificial Intelligence, 3202, 99-111, Springer, Berlin, Germany.
A. Appice, M. Ceci & D. Malerba. Mining Model Trees: a multi-relational approach. In T. Horvath and A. Yamamoto (Eds.), Inductive Logic Programming ILP’03 , Series: Lecture Notes in Artificial Intelligence, Vol. 2835, 4-21, Springer: Berlin. Szeged, Hungary September 29 - October 1, 2003.
M. Ceci, A. Appice & D. Malerba. Mr-SBC: a Multi-Relational Naive Bayes Classifier. In N. Lavrac, D. Gamberger, L. Todorovski & H Blockeel (Eds.), Knowledge Discovery in Databases PKDD 2003, Series: Lecture Notes in Artificial Intelligence, Vol. 2838, 95-106, Cavtat-Dubrovnik, September 22-26, 2003.
M. Ceci, A. Appice, D. Malerba & V. Colonna. Multi-relational Structural Bayesian Classifier. In A. Cappelli and F. Turini (Eds.), AI*IA 2003: Advances in Artificial Intelligence, Series: Lecture Notes in Artificial Intelligence, Vol. 2829, 250-261, Springer: Berlin. Pisa, September 23-26 2003.
A. Appice, M. Ceci, D. Malerba, D. Sacchi. Stepwise Model Tree Induction in a Multi-Relational Framework. Atti del XI Convegno Nazionale su Sistemi Evoluti per Basi di Dati, SEBD’03. 281-292. Cetraro, June 24-27, 2003.
M. Ceci. Naive Bayesian Learning from Structural Data, Ph.D. Thesis, Università degli Studi, Bari, Italy. (2005)
A. Appice. Learning Relational Model Trees Ph.D. Thesis, Università degli Studi, Bari, Italy. (2005)

Top of this page

FAQs

None yet available. Send all requests/comments to: Michelangelo Ceci, Annalisa Appice Dipartimento di Informatica, Università degli Studi di Bari (Italy).

Top of this page

Last Update: Tue July 19 2005 16:54:45 GMT+0200 (CEST)