Mining Matrix Pattern from Mobile Users - Distributed Artificial Intelligence, Agent Technology, and Collaborative Applications - page 245

Figure 10. Algorithm for defining multi-layer matrix
Matrix Create_Multi_Layer_Matrix(size1, size2, size3) {
s1 = size1;
s2 = size2;
s3 = size3;
M = Matrix.Generate(s1,s2,s3);
M.Initialize(0);
Return M;
}
Figure 11. Algorithm for collecting frequency counts
Void Collect_Freq_Counts(Position X, Position Y, Position Z) {
If Event().Status=True Then {
Single_Layer_Array(X, Y, Z)++;
Nconfidence = Frequency(X, Y, Z) / Window Size;
Single_Layer_Array(X, Y, Z).Value = Nconfidence;
}
}
logical domains are of interest to the mobile users. After the three variables are defined, the structure
of the multi-layer matrix is created. Figure 10 shows how the multi-layer matrix is created.
Step : Select Window Size for Each Layer
After the multi-layer matrix structure is defined, the next step is to select a window size for each layer.
It is possible for each window size in each layer to be different or to be of equal values. The onus rests
on the decision maker in terms of the amount of resources such as memory capacity and processing
power that they are ready to invest and the amount of accuracy or sensitivity the results needs to be.
The window size for each layer is then defined in relation to the reference of each layer that is the logi-
cal domain of each layer.
Step : Collect Frequency Counts for Each Layer
Figure 11 shows that after the window size is defined, the frequency count can start to be collected. It
is important to keep the frequency counts within the window size to prevent any waste of resources. It
is also essential for the data collection to be executed for a suitable timeframe of interest to the decision
maker. There is no point in gathering data during a wrong timeframe, such as gathering workplace ac-
tivity at midnight when workplace activities are well known to be minimal. Each frequency count will
contribute to the particular storage area of the matrix for later mobile user data mining purposes.
Step4:CalculateConidenceforEachLayer
Figure 12 shows that after the window size is defined and the frequencies are properly collected, the
initial requirement to calculate confidence is established. Confidence can now be calculated using the
formula: conidence = frequency / window size . The confidence is the ratio between frequency and