A PageRank-based collaborative filtering recommendation approach in digital libraries

Abstract

Original scientific paper In the current era of big data, the explosive growth of digital resources in Digital Libraries (DLs) has led to the serious information overload problem. This trend demands personalized recommendation approaches to provide DL users with digital resources specific to their individual needs. In this paper we present a personalized digital resource recommendation approach, which combines PageRank and Collaborative Filtering (CF) techniques in a unified framework for recommending right digital resources to an active user by generating and analyzing a time-aware network of both user relationships and resource relationships from historical usage data. To address the existing issues in DL deployment, including unstable user profiles, unstable digital resource features, data sparsity and cold start problem, this work adapts the personalized PageRank algorithm to rank the time-aware resource importance for more effective CF, by searching for associative links connecting both active user and his/her initially preferred resources. We further evaluate the performance of the proposed methodology through a case study relative to the traditional CF technique operating on the same historical usage data from a DL. Keywords: collaborative filtering; digital library; PageRank algorithm; recommendation approach; social networkPreporučeni pristup page-rank kolaborativnog filtriranja u digitalnim knjižnicama Izvorni znanstveni članak U sadašnje vrijeme opromnog broja podataka, eksplozivni porast digitalnih izvora u Digitalnim Knjižnicama -Digital Libraries (DLs) doveo je do ozbiljnog problema preopterećenja informacijama. Taj trend zahtijeva pristupe personaliziranih preporuka koji bi korisnike DL upoznali s digitalnim izvorima specifičnim za njihove individualne potrebe. U ovom radu predstavljamo personalizirani pristup preporuci digitalnog izvora koji kombinira tehnike PageRank i Collaborative Filtering (CF) u sjedinjenom okviru u svrhu preporuke odgovarajućih digitalnih izvora aktivnom korisniku generirajući i analizirajući mrežu u postojećem vremenu kako odnosa među korisnicima tako i odnosa među izvorima. Kako bi se obradila postojeća pitanja o postavljanju digitalnih knjižnica, uključujući nesigurne profile korisnika, nesigurna obilježja digitalnog izvora, oskudnost podataka i problem hladnog starta, ovaj rad adaptira personalizirani PageRank algoritam kako bi rangirao važnost izvora koji vodi računa o vremenu učinkovitijim CF, tražeći asocijativne linkove koji povezuju i aktivnog korisnika i njegove/njezine početno preferirane izvore. Također ocijenjujemo performansu predložene metodologije kroz analizu slučaja vezanog za tradicionalnu CF tehniku koja koristi iste podatke iz Digitalne knjižnice.Ključne riječi: digitalna knjižnica; društvena mreža; kolaborativno filtriranje; PageRank algoritam; pristup koji se preporučuje

Massive-Scale Data Mining to Enhance Digital Library with Applications in College Education

“…Digital libraries have foreseen the academic resorts that users can choose [ 1 – 10 ], but with the explosive effect of academic resources, the phenomenon of digital library information overload has become more and more serious [ 11 ], and finding a suitable academic resort has become a huge problem [ 11 , 12 ] and using recommender system support to address this problem [ 13 ]. With the help of the recommendation system, the relevant intelligence is “pushed” to the digital library users so that they can keep more benefit [ 10 ].…”

“…In the same year, the American Association for Artificial Intelligence (AAAI) organized a special conference to focus on the development of recommender systems [ 15 ]. Around 2000, Cornell University Library took the lead in breaking through the relatively mature library personalized office system My Library@Cornell system and provided a purpose model for many subsequent digital library witness systems [ 11 ]. Since then, the personalized witness utility of the digital library has gradually improved.…”

Time-Bin-Based Neighbourhood Algorithm for Temporal Effects in Recommendation Systems

“…Collaborative filtering (CF) is a commonly used RS implementation technique in the literature [1][2][3]. From the application perspective, movie-based RS is a favourite research subject in CF [4][5][6].…”