Towards an Agent-Based Knowledge Management System - Developing Concepts in Applied Intelligence - page 98

We have used PL-SQL developer Ver. 6.0.3.893. a s our Multi-Agent development
tool for developing TA, RA , LA and KCA . Reader may refer to
http://www.ju.edu.jo/sites/Academic/obein/Lists/Academic%20Conferences/AllItems.a
spx for the pseudo-code of the agents. . We have run several experiments where each is
designed to measure a specific functionality as follows:
1. Test-Retest Reliability: The reliability of the model depends on reliability of the
agents. We have measured the reliability of TA, RA, KCA and LA .
2. Measuring Internal Consistency Reliability for RA : We measure internal consis-
tency reliability for RA with regard to a single parameter. Figure 4 shows the val-
ues of internal consistency reliability for 5 trials. The minimum value is 80%.
3. Measuring System Efficiency: We have run experiments to measure the efficiency
of RA and KCA. We increase the number of cases and experts in the following or-
der (1000, 2000, …, 6000) for the case-base and (45, 90, 135, 180, 225, 270) for
KB. Figure 5 shows the additional execution times RA needed to investigate the
new cases. Figure 6 shows that the performance RA varies in a stable manner when
the number of cases of additional organizations increases. (cf. Figure 7).

Internal Consistancy Reliability of RA With Regard to Course-Name

Execution Time of RA among Number of Organization

120

3.24

3.22

100

3.2

80 3.18

3.16

60

3.14

40 3.12

3.1

20

3.08

3.06

0

1 2 3 4 5 6

1 2 3 4 5

M inor Adjustment Trials Organization Levels

Fig. 4. Measuring the internal consistency Fig. 5. Execution times that RA needs
reliability for RA. to investigate new cases.

Average Mean
To t a l Mea n

Variance betwe en Values of Column (B) and the Average Mean Variance Comparis on Re garding Ave rage Me an and Total M e an

0.12

0.06

0.1

0.04

0.08

0.02

0.06

0.04

0

0.02

1 2 3 4 5 6

-0.02

0

-0.02 1 2 3 4 5 6

-0.04

-0.04

-0.06

-0.06

-0.08

-0.08

Organization levels

Organization Level

Fig. 6. Variance values of execution vs Fig. 7. Variance to average mean vs
Org. levels & average mean. variance to total mean.

The Averagre Effect of Five Course_Name_Weight Values on RA
Extracte d Re sults

12

10

8

6

4

2

0

25 20 15 10 5 0

Number of Ext ract ed Cases

Fig. 8. Extracts by RA as course_name increases.