Improved Link Prediction Using PCA

Abstract

Link Prediction is known as a challenging problem in the area of online social media. Earlier, learning model for link prediction task has been proposed by many researchers. But the classification of imbalanced and high dimensional data is an interesting and challenging problem in machine learning due to presence of unbalanced and redundant or correlated data which break down the classification performance. In this paper, we have balanced the data and used Principle Component Analysis (PCA) to reduce the correlated data and improved the performance of link prediction model. Experiment is carried out on social network data set and the use of PCA method has improved the performance in classification of links.