SC740 SEMINAR REPORT 03 for Frederick L. Jones

PRESENTER: Dr. Maria Cobb

OVERVIEW

Dr. Maria Cobb first described the primary database that the Mapping Sciences Section of the Naval Research Laboratory, Stennis Space Center uses, i.e., the Geospatial Information Database (GIDB). The GIDB was developed using an Object Oriented approach. The GIDB was described in terms of its Spatial Data Interface Components. Dr. Cobb then described Spatial Data Mining, Data Mining in general and Object Oriented Database Mining.

Dr. Maria Cobb then focused on Fuzzy Logic techniques for implementing spatial data mining of the GIDB. Dr. Cobb discussed a specific model, the Abstract Graph (ASG) Model, for integrating a fuzzy set model for spatial relationships. Then Dr. Cobb focused on of exactly how to integrate Object Oriented (OO) Data Models and Fuzzy Logic ASG Data Models.

THE GIDB SPATIAL DATA INTERFACE COMPONENTS

The GIDB was described by Dr. Maria Cobb by its spatial data interface components. These spatial data interface components are the following: (1)Data (2)Integration (3)Data Analysis and (4)Visualization. In the Data component the data in the GIDB varies from video and audio to longitude and latitude coordinates and their associated topology and geometry. In the Integration component the data is included in a DBMS, warehoused and integrated into the GIDB database. In the Data Analysis component the data are made available for decision support, for exploration and for interpretation. In the Visualization component the data are presented in either 2D or 3D detailed graphics.

SPATIAL DATA MINING AND DATA MINING IN GENERAL

Since the GIDB is a Spatial Database, Dr. Maria Cobb explained Spatial Data Mining. Dr. Cobb referenced Kaperski’s three primitives of Spatial Data Mining, (i.e., the spatial characteristics rule, the spatial discriminant rule and the spatial association rule), to show that Spatial Data Mining involves both explicit information, as well as intrinsic relationships.

Data Mining in general was described as a knowledge discovery algorithm. This knowledge discovery algorithm involves the following: (1)clustering, (2)spatial association exploration, (3)approximation and (4)aggregation.

Since the GIDB was developed with an Object Oriented approach, Object Oriented Database Mining was also discussed by Dr. Maria Cobb. OODB Mining involves the following concepts: (1)class hierarchies and concept hierarchies, (2)object identifiers and indexing schemes, (3)method sets and associated rules and (4)composite objects and association rules.

FUZZY SPATIAL RELATIONSHIPS AND THE ABSTRACT SPATIAL GRAPH MODEL

Fuzzy spatial relationships were defined by Dr. Cobb by the following characteristics: (1) they are Minimum Bounding Rectangle (MBR) based, (2)their Topological relationship definition is bi-level, i.e., (a)it is position [r_x,r_y] based on Allen’s relations and (b) it is also linguistic, since it is based on fuzzy sets, and (3)their Directional relationship definitions are also bi-level.

The Abstract Spatial Graph (ASG) Model makes use of fuzzy set membership classes and their degree of set membership to assign object subgroup node weights as well as area weights in an ASG. (I assume that the ASG is a weighted graph with both arc as well as node weights. Because of these fuzzy set degree of set membership weights being used in the ASG, queries such as the following can be made: Is Object B somewhat north of object A?

INTEGRATING OBJECT ORIENTED AND ASG DATA MODELS

Dr. Cobb’s final topic was a explanation of exactly how to integrate Object Oriented (OO) Data Models and Fuzzy Logic ASG Data Models. Dr. Cobb explained the compatabilities of the two data models at the Geographic feature level. She also explained the Spatial Data Access method implications of the integration of the two data models. Dr. Cobb concluded her presentation by stating that Object Oriented and Spatial Data Mining techniques are compatible, and thus can be integrated.

SUMMARY AND CONCLUSIONS

Dr. Maria Cobb described the primary database used by the Mapping Sciences Section, NRL, Stennis Space Center, i.e., the GIBD. The GIDB was developed using an Object Oriented approach. The GIDB was described in terms of its Spatial Data Interface Components. Dr. Cobb then described Spatial Data Mining, Data Mining in general and Object Oriented Database Mining.

Dr. Maria Cobb then focused on Fuzzy Logic techniques for implementing spatial data mining of the GIDB. Dr. Cobb discussed the ASG Model for integrating a fuzzy set model for spatial relationships. Then Dr. Cobb focused on of exactly how to integrate Object Oriented (OO) Data Models and Fuzzy Logic ASG Data Models. Dr. Cobb concluded her presentation by stating that Object Oriented and Spatial Data Mining techniques are compatible, and thus can be integrated. I also concur with her conclusions.