Dataset Files
Find the dataset files required for the algorithm here
Dataset Description
We use Epinions' dataset to test our algorithm. Epinions contains information about users and the reviews that they make about products. Those users form a network, as they declare their trust or distrust with each other. When a user trusts another user, then there is an edge with label 1 between them. If there is distrust among them, then the label of the edge is -1.
Since we did not find a complete Epinions dataset with both information about reviews and negative links between users, we combined two different datasets found online. The first one is a mysql relational database containing information about the reviews, the users and the network between those users. The problem though, was that this network contained only positive edges. Then came the need for a second one , which is a directed signed network for Epinions' users, with both positive and negative edges. Using Epinions' unique user identifier, we created and use a dataset, that contains the network from the second dataset and the reviews from the first one only for users that participate in the aforementioned network.
The created dataset contains 31.322 experts(users), 587 skills(categories) and in our network there are 210.078 edges between the experts. From the 210.078 edges, the 35.150 are negative (circa 16,7%).