Information Technology Reference
In-Depth Information
2 Graph Model for Pattern Recognition
The overall approach of this algorithm begins with the identification of a set
of relevant keywords. Once these are selected, we then aggregate the relative
distances of the keywords with a document.Thisinturnisusedtoconstruct
a weighted directed multigraph that generates representing vectors for each
document in a high dimensional feature space. These vectors can then be
used to determine similarity values for any pair of documents.
2.1 Summary of Our Method
Step 1:
Using a weighted directed multigraph to find a signature vector for
each document.
Step 2:
Calculate the similarities between any two documents via their sig-
nature vectors.
Step 3:
Using Quasi-Clique Merge clustering method to classify all
documents.
We will explain the details of each step by a simple example.
2.2 Details of the Step 1
To have a clear view of the algorithm, we will use the example illustrated in
Figure 1 [26] to explain the procedure.
Fig. 1
A fraudulent email
