Link prediction illustrative image. Source
I worked under the IIMAS's Department of Systems Engineering and Automatization to recreate and afterwards improve upon previous results in predicting future academic collaborations between authors using only associated topological data. This problem was analyzed in Link Prediction using Supervised Learning (Hasan et al., 2006), in which the authors presented results with the following models:
•Decision Trees
• SVM with Linear and RBF Kernels
• K-Neighbors
•Gaussian Naive Bayes
•Multi-Layer Perceptron
•RBF Network
•Bagging on the previous models
After assuring the replicability of the original work, I improved upon its results by creating a reduced feature vector for a given author using SVD on the collaboration network's adjacency matrix.
The repository for this project can be found here.
