Improving random walk rankings with feature selection and imputation

TL;DR

This paper presents a model that improves random walk rankings in a semantic network by applying feature selection and imputation, achieving second place in the Science4cast Competition.

Contribution

It introduces a novel approach combining feature selection and imputation to enhance random walk-based link prediction in semantic networks.

Findings

Achieved a test score of 0.92738, ranking second in the competition.

Demonstrated that feature selection and imputation improve ranking performance.

Model variations were analyzed for their impact on results.

Abstract

The Science4cast Competition consists of predicting new links in a semantic network, with each node representing a concept and each edge representing a link proposed by a paper relating two concepts. This network contains information from 1994-2017, with a discretization of days (which represents the publication date of the underlying papers). Team Hash Brown's final submission, \emph{ee5a}, achieved a score of 0.92738 on the test set. Our team's score ranks \emph{second place}, 0.01 below the winner's score. This paper details our model, its intuition, and the performance of its variations in the test set.