- Example 1: User Similarity in a Social Network without Edge Weight
- Example 2: User Similarity in a Social Network with Edge Weight
- Example 3: User Similarity and Product Recommendation
- Example 4: Using the Sources and Targets Tables as Inputs
Examples 1 and 2 analyze a social network of users (for example, in an application such as Twitter) as shown in the following figure and their relationships as followers and leaders based on the 'likes' each user gets from others.
The preceding figure is converted to the following figure, which is a bipartite representation of the same network. The nodes on the left side of the following figure are source vertices, or hubs, and constitute the 'followers' column in Table bbb. The nodes on the right side of the following figure are the target vertices, or authorities, and constitute the 'leaders' column in Table bbb.
The pSALSA algorithm assigns scores to both sides. In the output table (Output), the hub_followers column shows similar users for each follower based on the hub_score. Likewise based on the authority_score, leaders who are close to followers are output in the authority_leader column. A higher score indicates greater similarity. Typically the authority_score is interpreted as user recommendations (in this case the closer leader) and the hub_score is interpreted as user similarity.