Portale per Data Mining: Hardware

1

A WEB Based Cooperative System for Pollution Data
Analysis and Environment Healt Monitoring
A Research and Development Initiative Promoted by
DI.S.T.A.
Funded by EEC in the framework of BEEP
project and by MURST in the framework of MITILUS
project
Project Management Attilio Giordana (Computer
Science) Aldo Viarengo (Biology) Team
M. Botta, Burlando, L. Portinale, A. Serra, M.
Rapetti, G. Porcelli.
2
Project Goals
To provide a user friendly apparatus world wide
accessible for sharing and analyzing biological
data collected by the environment monitoring
laboratories
To develop new data mining algorithms oriented to
biological data analysis.
3
Hardware Support Web serverDB serverBeowulf
(20 Pentium III 800)
Beowulf (20 PC)
4
User Environment
5
Software Architecture
Data Storage (192 Gbytes)
WEB Server
Oracle Database Manager
Servelets
Java Interface
Data Mining Algorithms
6
A User Fiendly Graphic Interface
........
Dataset name
Tool1
Tool2
Tool3
Tool4
Tool5
op1
op2
op3
view 1
op1
op2
op3
op4
view 2
..............................
7
Data Intensive AlgorithmsRun in Parallel on the
Beowulf
Algorimi ad uso interattivo
D-Tree
Algoritm Server
Neural Net
Servlets
..............
G-net
Sequence Analysis
Cluster
8
Workpackages
WP1 Database Design Meta data
WP2 Graphic Interface design and implementation
2.1 Approach selection 2.2
User autentication procedure 2.3 Oracle
interafce 2.4 Data visualization
WP3 Tool configuration interface 3.1
G-Net 3.2 Mine-Rule 3.3
Clustering Algorithms 3.4
Characterization Algorithm 3.5
Decision/Regression Trees 3.6 Neural
Networks
9
Workpackages.....
WP4 Servlet implementation
WP5 Algorithm server implementation 5.1
Design 5.2 implementation
WP6 Algorithm implementation 6.1 New
KDD algorithm implementation 6.2
Existing algorithm revision
10
Work-Flow
15/1/2001
15/2/2001
15/3/2001
Wp1
Wp1 (revision)
Wp2.1
Wp2.2 - Wp2.4
Wp3.1-Wp3.6
Wp4
Wp5.1
Wp5.2
Wp6.1 - Wp6.2