AMiner Citation-Data Preprocessing for Recommender Systems on Scientific Publications

Abstract

Recommender Systems (RS) are used to find user's interested items among a huge amount of digital information, recently called Big Data, with the purpose of making valuable personalized recommendations. These systems use data from digital, online libraries to train, test and evaluate system's efficiency. Along this line, data preprocessing is an essential and valuable step to achieve information-preserving data reduction and, in addition, to create input files with the appropriate format needed by a RS. This paper describes our approach for data preprocessing using a scientific publications' dataset (Computer Science) found in AMiner (https://www.aminer.org/). The proposed approach consists of two phases: creation of a collection of articles based on user preferences and preprocessing this collection. The experimental results demonstrate the value of our approach with at least 79.8% information-preserving data reduction. © 2021 ACM.