Visualizing Access Logs of a Scientific Digital Library Effectively as Multiple Time Series Using Modified Horizon Graphs
Trung Hieu Le, Tran Khanh Dang, Tran Tri Dang, The Nhan Luong · 2019
Libraries usually subscribe to digital data sources. Their members (students, teachers, researchers, ...) gain access to resources stored in these external services via proxy servers managed by the local libraries. Library managers have the need to know the usage of subscribed services by members so that they can make decisions, for example, whether to renew a subscription or not. Proxy servers access logs store valuable information for them, however, it's difficult to get the answers directly from log files because the raw data is unstructured and massive. In this paper, we describe a two-step technique to support library managers in viewing and understanding usage information more effectively using the data provided by proxy servers' log files. The first step consists of preprocessing access logs from proxy servers and constructing multiple time series from the data. A modified version of horizon graphs was developed in the second step to visualize many time series on one screen. The screen space is divided unequally to different time series based on their relative importance with each other. In addition, an interaction method and a dynamic legend are also introduced to give a better understanding of the graphic representation. The proposed method is implemented as an application running in a web browser. Experiment results with real world data gathered in one year show that the proposed technique can improve users' performance on data exploration and analysis tasks notably.