Evaluating the Influence of Human and Non-Human Web Traffic on Academic Repository Analytics

Aguilar-Bueno J.; Alvarado-Lugo R.A.; Lujan-Leon J.P. · C3 2025 - IEEE Colombian Caribbean Conference (2025) · DOI: 10.1109/c366505.2025.11340327

Resumen

Institutional Repositories (IRs) play a critical role in the dissemination of open-access knowledge, yet their performance evaluation is often hindered by non-human web traffic, such as bots and crawlers, which distort key performance indicators (KPIs). In this study, we present a lightweight data analysis methodology designed to identify and filter non-human activity in IRs. Using nearly one year of real data, we combined serverside statistics from DSpace and Solr with anomaly detection techniques, including Tukey's rule and time-interval analysis, to distinguish human-driven downloads from automated traffic. Our results reveal that approximately 89% of the recorded traffic is non-human, significantly inflating raw KPIs such as downloads and views. In addition, we found massive non-human web traffic originating from only 0.01 % of IP addresses, while 99 % of human users performed up to 8 downloads per day. The proposed approach reduces these distortions and provides an accurate assessment of IRs.