Mining Matrix Pattern from Mobile Users - Distributed Artificial Intelligence, Agent Technology, and Collaborative Applications - page 246

Figure 12. Algorithm for calculating confidence
Confidence Calc_Confidence (Frequency, Window Size)
Frequency_Valid = (Frequency > Min && Frequency < Max);
Window_Valid = (Window > Min && Window < Max);
If (Frequency_Valid && Window_Valid) Then {
Confidence = Frequency / Window Size;
If (Confidence < Confidence_Threshold) {
Confidence = 0;
}
Else {
Location.Mark();
Return Confidence;
}
}
}
Figure 13. Algorithm for calculation of relationship confidence
Confidence Calc_Relationship_Conf (Conf1, Conf2)
C1OK = Conf1.validate();
C2OK = Conf2.validate();
If (C1OK && C2OK) {
Confidence = Average(Conf1, Conf2);
If (Confidence < Confidence_Threshold) {
Confidence = 0;
Return Confidence;
} Else {
Conf1.Location.Mark();
Conf2.Location.Mark();
Conf_Table.Set(Conf1, Conf2, Confidence);
}
}
}
window size. Therefore, both frequency and window size will affect confidence and only window size
is under the control of the decision maker. Therefore the adjustment of window size will have a direct
effect toward the confidence value.
Step5:SelectConidenceThresholdforEachLayer
After the confidences of all the layers are determined, it is now time to filter out the non-significant
confidence static nodes. To start with, the confidence threshold from each layer is determined, obtained,
and compared against each storage area of the matrix in every single layer. The confidence threshold
can be different in each layer, and all those storage areas with the confidence lower than the confidence
threshold will be marked insignificant and will not be taken into future considerations. The rest, which
are highly significant confidence static nodes will be identified and used in the next step.