Upload
jasmine-collins
View
213
Download
0
Embed Size (px)
Citation preview
Intelligent Information Expert System for Employment
and General Purposed Fuzzy Shell
2
Technological centre:
Solware Information Technology Ltd.Tasks:
• coordinator• project leader• software development
Knowledge centre:
Budapest University of Technology and Economics, Department of Telecommunication and Telematics Tasks:
• fuzzy algorithms• aggregation
- general searching-offering system
- general fuzzy shell
Members of the Consortium
3
• Project goals
• Background
- expert systems
- fuzzy definitions
- fuzzy rule-based expert systems (fuzzy shell)
- comparison• Project
- fuzzy based searching-offering systems
- employment expert system• accomplishment with modern IT methods• design, optimisation system parameters
Conclusions
Agenda
4
Goal: Development of a specific and a general purposed fuzzy rule-based expert system
Two steps:
1. Development of a fuzzy-based searching-offering system2. Development of a general purposed fuzzy-shell
Mile stones:a. Fuzzy-based searching-offering subsystemb. Job searching subsystemc. Applicant searching subsystemd. Fuzzy shelle. Optimisation the parameters of employment (job and
applicant searching) f. Design fuzzy-shell demo program
Goal
Project Goals, Tasks
5
Knowledge base
Antecedence Consequence
Conclusion
Character of the knowledge base:
The rules are applied to crisp values and intervals
Difficulties:
• very large knowledge base, too many rules
• the uncertainties are handled not efficiently
• inflexible system: no applicable rule no result
Backgr
.Expert Systems
6
• Fuzzy setA is a set on the X universe,
Fuzzy set: belongs to the given A set so that the measure of this membership is not 1 or 0 (x belongs to A or not) but a value between the two
• Membership function
The measure of the belonging
• Fuzzy logicGeneralisation of the two-valued Boole type logic
income
(USD)
Backgr
.
1,0Xx
A
1
0
A
3000 USD
A=more then 3000 USD
Fuzzy Definitions
XA
Xx
7
Observation Crisp output
Rule-base
Backgr
.
Fuzzyfication
unit
Defuzzyfication unit
Inference engine
Inference Algorithm i.e.
Mamdani Sugeno, etc..
If Then
Fuzzy Rule-Based Expert Systems (Fuzzy Shell)
8
Advantage over the classical expert systems:
• less rule • decrease computational complexity
c = decreasing factor against symbolical expert systems
• good handling the uncertainties• robust system (overlapping rules)
Disadvantage:• decreasing accuracy
Applications have great perspectives on the areas where the uncertainty is large and not needed very accurate result
Additional new components comparing to other (fuzzy) expert systems:
• build in interpolative methods
• hierarchical systems
Backgr
.
kk
cTT
Comparison
9
Problems by finding the partners each other:
- in discrete case:search
offer
Similarity
matrix
Projec
t
1
0
Offer Search
0.45
A solution: using fuzzy sets- in continuos case:
• uncertainties:
the searching partner
doesn’t know exactly
what he/she wants
• weight of viewpoints
are differences and can be
changing during the process
Fuzzy Based Searching-Offering Systems
10
• Variables were chosen and structured by professional employment experts
• Typical variable groups- income (salary and other)
- personal skills (education, language etc.)- workplace information (distance, firm size)
• More problematical case were handled:
- distance: - taking into account the infrastructure the system able to calculate the distance in time
- branches: - all the branches are covered by similarity matrix
- weight: - the weights of the variables can be iterated after the analysis of the output
Projec
tEmployment Expert System
11
Algorithms
- user surface on web and windows environment
- SQL database
- XML, MTS applications
Projec
t
Job searchinguser surface Employment
agency surface
Job offering user surface
User surfaces
DatabaseApplication logic
Fuzzy system
Accomplishment
12
Architecture of Employment Expert SystemProjec
t
Internet client
Internet client
Internet client
Web server Application server
XML configuration
files
SQL data base
13
• optimisation on real data
• the learning algorithm is a type of evolutionary algorithm: bacterial algorithm
Optimisation with learning methods
Check the results on test set
Projec
t
Default parameters(employment experts)
Choose the parameters for optimisation
Design, Optimisation of System Parameters
14
Summ
a
• The method- Advantages and disadvantages of fuzzy-based expert systems
- Motivation of using fuzzy methods
• Until nowGeneral searching-offering system
right now: Job searching subsystem
• Future - Applicant searching subsystem
- General purposes fuzzy shell
Conclusions