Comparison of Apriori and Top-Down Mining Algorithms on Co-Authors’ Relationship in DBLP Online Bibliography

ABSTRACT

Academic research remains one of the effective ways for advancing humans’ status of living and the powerful tool for promoting intelligence. However, diversity comes in human due to various reasons such as age, experience, accessibility; thus, making their research strengths and outputs to vary, accordingly. Nonetheless, collaborative and cooperative relationship was identified as an important way to facilitate genuine ideas creation, synergistic output that would have been impossible otherwise. Furthermore, it increases research outputs both in terms of quantity and in quality. Amongst the limited studies that focused on efficient co-author relationships mining, Apriori and top-down algorithms are widely used techniques. In this study, both methods were compared for frequency and association mining in order to extract some inherent virtues amongst authors in DBLP, a public research repository. Correlation among authors is strongly assumed to be equivalent to how often similar authors appear together in research publication. Implementation of algorithms was studied carefully on a computer system with Intel Core 2 Quad processor (2.40GHz each), 2 GB RAM, and 160GB HDD. The program was characterized on a Windows 7 and WAMP server. Results of experimental study on a total of 26,942 authors that were uniquely extracted from 8,482 bibliography records show that top-down performs better than Apriori Algorithmic approach.

INDEX TERMS: Apriori, Co-Authors’ Relationship, Data Mining, DBLP, Top-down Mining