Sign Prediction on Social Networks Based Nodal Features

Table 1

Algorithm of pseudo-code of -model.

---

Input: Network adjacent matrix.

Initialization: For each node , , , , .

---

For each edge , do the following 5 steps:

Step 1. Compute the neighbor set of each node.

  If , , ;

  If , , .

Step 2. Compute the number of special quadrilaterals.

  For %The existing sign of is not considered.

  For %The existing sign of is not considered.

  If , set  = +1,

where is the number of quadrangles in Figure 2(a).

---

  Similarly, compute , , , .

---

Step 3. Compute the similarity and dissimilarity.

  .

  Similarly, compute , and .

Step 4. Compute values.

  .

Step 5. The sign of each edge is predicted.

  Set , .

  If , ;

  If , .

  If and , set ;

  If and , set ,

where is the proportion of positive links in the network.

Output: The sign of each edge .

---