note: the focus is on the performance and design of three text classifiers

regarding working of multiple intelligent agents in an autonomous software installation/uninstallation environment.

The above mentioned agent based uninstall/installation process will have following sequence of actions (for multiple agents) , agents will work in following sequence:

1. An interface for communication between with the user and the program
2. Deploying agents to conduct(install/uninstall) task on the machine
3. accessing the installation/un installation rights.
5. Checking the requirements/physical constraints (available space and software etc).
6. Initiating the uninstall/installation process using agents.
7. Smooth and successful continuation of uninstall/installation process (analyzing the text and responding to that), the main function
· Analyzing the text of every uninstall/installation window/screen based upon three text classifiers (naive baesian, bayesian belief and svm)
1) naive Bayesian model/classifier
2) Bayesian belief network model/classifier
3) SVM based classifier
· One by one action selection / decision making on the base of above mentioned text classifiers.
In case of an unknown environment or unknown text/case, capability of dealing with the situation (learning capability) and adding that to the classifiers knowledge base
Communication and collaboration capabilities along with learning
capabilities (unknown environments) .
· Updating the directories (system) after installation/uninstallation is complete and completion msg should be displayed
· Storing and Comparing the results of all the models at the end using a database and making a comparison of them all to show which one is the better
8. change in environment(directories) upon successful completion of the uninstall/installation process.
9. Reporting the success of the process and displaying the results using an interface.

just for a single machine, i donot want it for a networked/distribute environment.
Hav eyou worked in jade before?

you will have to provide me details regarding the model implementation (naive baesian, bayesian belief and svm) for my design documentation.
one more thing, you will have to send me some basic work as phase 1 on 27 for my preliminary search.


i actually wanted the same, assess , analyze and then commit so that no chance of failure, like my previous experiences.
thankyou

--- Update ---

plz start with interface and text classifier design, as reading from windows could be deal afterwards and it is the most difficult task
main design that is to submitted on 10 is of 3 models, you can apply them on a text file that has installation text copied into that along with output options