Mining Matrix Pattern from Mobile Users - Distributed Artificial Intelligence, Agent Technology, and Collaborative Applications - page 248

Figure 14 shows that multi-layer relationship confidence can be calculated by using the formula:
multi-layer relationship confidence = (Relationship Confidence Layer 1 + Relationship Confidence
Layer 2 + … + Relationship Confidence Layer N) / N . It is essential to note that the number of domains
that can be selected is limited to the number of layers, while it can be at least one per size of the layers.
It is the average of the relationship confidence in a particular static node over different layers of logical
domain.
Step9:SelectMulti-LayerRelationshipConidenceThreshold
After the multi-layer relationship confidence is obtained, it is now time to test whether the multi-layer
relationships are significant or not. The decision maker can test it by means of determining a multi-layer
relationship confidence threshold. By a careful selection of a multi-layer relationship confidence thresh-
old, all non-significant multi-layer relationships can be discarded. This is determined by the multi-layer
relationship confidence lesser than the multi-layer relationship threshold.
It is essential that the decision maker be aware that for the multi-layer relationship confidence to
come to this stage, it has to pass two previous tests for two different stages. Therefore, these multi-
layer relationships are most likely highly significant and only very few are discarded with a population
sample of very large size.
Step 0: Generate Multi-Layer Matrix Pattern
Figure 15 shows that after all the confidences, the individual static node confidence, the inter-static
node relationship confidence, the multi-layer relationship confidence calculated and screened, it is time
to display the result to the users. The system should be adjusted in such a manner that by this stage,
the amount of knowledge found must be of high quality and the volume of knowledge must be within
an acceptable level that does not overload the decision maker. The ability for information loading is
dependent on the decision maker and is another area of research.
Perfor MAnCe eVALuATIon S
The purpose of matrix pattern is to provide an efficient way for obtaining useful knowledge from mobile
users. By using static nodes as the main domain for mining, it will be more cost efficient compared
Figure 15. Algorithm for generation of multi-layer matrix pattern
Void FindPattern (Multi Layer Matrix M) {
For (I = 0, M.Size(), M++) {
If M.I.Significance = True {
Display M.I;
Display M.I.Confidence();
}
}
}