|
  
Next:Model Generation Up:Framework Previous:Model GenerationVerification and
System Configuration
The process of model generation, validation, and revision requires that user closely interacts with the system to construct, execute, and revise the queries. A web-based system will have the additional
advantage of allowing the user to remotely validate the model across the Internet. Figure 3 shows the
architecture of the proposed implementation. This architecture consists of Java clients, an HTTP server, the mining engine, database management system (e.g., IBM DB2), an index and image archive, and a
library contains various feature extraction, template matching, clustering and classification modules: 
Figure 3: Structure of the environmental model at the cell level.
- Java Clients
: Each client allows the user to hypothesize, validate, and revise epidemic models during the model construction phase. Once the model is verified against a small collection of data,
the client visualizes the results of applying the model to larger geographic regions.
- Model-Based Mining Engine
: The model-based mining engine accepts requests from the clients to construct, validate, and revise the model. The model submitted from the client is parsed into a
set of operands and operators. For each request from the client, the corresponding mining operator is invoked from the mining library against the data and/or indices stored in the data and
index archive. Furthermore, the engine is also responsible for revising the model after each iteration by the user.
- Database
: All metadata, such as case records, patient data, and image descriptors, are stored in a relational or object-relational database such as IBM DB2.
- Image Archive
: The images are stored in a virtual file system storage hierarchy that includes both disk and tertiary storage.
- Mining Library
: The software library contains sets of mining operators that perform classification, clustering, and other data cleansing and processing functions.
Note that the proposed architecture shares a number of features with generic database systems. Each hypothesized model corresponds to a query, and the validation of a model corresponds to the
processing of the query. However, this framework also supports model revision, a feature not generally available in conventional databases.
Next:
Model Generation Up:Framework Previous:
Model GenerationVerification and |
