Upload
maurice-sutton
View
213
Download
1
Embed Size (px)
Citation preview
A hybrid-expert-system based tool for scheduling and decision support
F.Franek, McMaster University+Terren Corp., Canada
V.L.Rosicky, Terren Corp., Canada
I.Bruha, McMaster University
Presenting TESS (Terren Expert System Shell), tightly coupled with the relational database back-end, used for scheduling of booking actions and decision support of marketing activities in GREENWICHTM software system for tour operators.
Slide 1
Need: most of tour operators in North America operate in similar manner and very slim profit margin. Hence need for some “standard” software for tour operators.
Problem: despite similarities, certain differences in the sequence of booking actions across the whole business. Not only different actions are taken, but there timing and sequencing may be different.
Solution: kind of “standard” software that may be easily tailored to the individual needs of each operator and that can produce action plans based on the booking data accordingly.
Abstract problem: easy reconfiguration based on some rules, action scheduling (planning) based on some rules.
Implementation: Rule-based expert system TESS!Slide 2
Why TESS: Very flexible, very expressive
r1: (0.3*P(x,”abc”) [>=.3] & ~Q(x,y)) | R(2,3.4,y) ==and:f1,or:f2=> 0.7*~S(y,x) [>.6]
Why not TESS: Too flexible, too expressive, no “native” database access.
Could not let users (of all people!!) to do it on their own, needed to provide natural and integral access to database.
Solution: Design a language that limits the expressiveness and flexibility, but allows “natural” expressions concerning database access!! SQL!!!
Slide 3
SELECT DepositWarning FROM Tess WHERE
AND booking_number=@bkg_number
AND agency_office=@ao_id
AND trip_id=@ts_id
AND trip_departure_date=@departute_date
AND BookingCompleted
AND ConfirmationLetterSent
AND (SELECT deposit FROM booking WHERE
booking_number=@bkg_number) = 0
AND (SELECT deposit_date FROM booking WHERE
booking_number=@bkg_number) >= @todaySlide 4
TESS Rules
TESS1
temporary
Relational Database
DataData
TESS native rules
TESS2
compiled TESS rule base
temporary
ACTIVITY ENGINE TESS INFERENCE ENG.
Actions
+
Activities
DataData
Schedule (Plan) of Actions+Activities
Slide 5
TESS Rules
Relational Database
SQL QUERY BUILDER
Rule Editor
Additional Advantage
How is the scheduling done?
Day-to-day simulation of actions to produce the plan of actions. The reason: the combinatorics of mutual interactions is not conducive to analytical solution. Despite of this, the natural constraints limit the combinatorial explosion to tractable level. Slide 6
Slide 7
CONCLUSION
For booking actions two-valued (1,0) logic suffices. Not so for marketing activities, fuzzy-logic or other uncertainty-handling approach necessary.
TESS natural for fuzzy-logic implemented through Certainty Value Propagation Functions and and or. In the setting of marketing activities we ran into the usual problem - where do the numbers come from? - so we are experimenting with neural nets in the place of CVPF’s and their training.