Web Usage Analysis: New Science Indicators and Co-usage

Xavier Polanco, Ivana Roche, Dominique Besagni
{polanco,roche,besagni}@inist.fr

Institut de l’Information Scientifique et Technique (INIST / CNRS)
2 allée du Parc de Brabois – 54514 Vandoeuvre-lès-Nancy – France

Mots clés :
webométrie, fouille de données d’usage du Web, indicateurs de la science, comportement utilisateur Web, analyse co-usage, serveur Web Keywords:
webometrics, Web usage mining, science indicators, Web user behaviour, co-usage analysis, Web server Palabras clave:
webometría, minería utilización de la Web, indicadores de la ciencia, comportamiento utilizador de la Web, análisis co-utilización, servidor Web

Résumé

A new type of statistical analysis of the science and technical information (STI) in the Web context is produced. We propose a set of indicators about Web users, visualized bibliographic records, and e- commercial transactions. In addition, we introduce two Web usage factors. Finally, we give an overview of the co-usage analysis. For these tasks, we introduce a computer based system, called Miri@d, which produces descriptive statistical information about the Web users’ searching behaviour, and what is effectively used from a free access digital bibliographical database. The system is conceived as a server of statistical data which are carried out beforehand, and as an interactive server for online statistical work. The results will be made available to analysts, who can use this descriptive statistical information as raw data for their indicator design tasks, and as input for multivariate data analysis, clustering analysis, and mapping. Managers also can exploit the results in order to improve management and decision-making.

1 Introduction

Two scientific communities are dealing with Web analysis related questions. This is the reason why we can observe in the literature two traditions about the analysis of the Web. One developed by people coming from documentation, and the other by computer scientists. The first was developed in the field of information science under the appellations of “webometrics” (Almind & Ingwersen, 1997), or “cybermetrics” (cf. http://www.cindoc.csic.es/cybermetrics ) while seeking to extend the informetric techniques to the analysis of the Web (Björneborn & Ingwersen, 2001; Ingwersen & Björneborn, 2004). The second one arose in the field of the computer science while seeking to extend the data mining techniques to Web analysis under the appellation of “Web mining” (Chakrabarti, 2003) and according to three main categories: Web structure mining, Web content mining, and Web usage mining (Kosala & Blockeel, 2000). We work at the border of these two traditions: we consider informetrics from the point of view of computer-based technologies. The Web represents a new environment for the quantitative studies of science, and a new family of computer-based science indicators can be developed. This article deals with a system able to produce descriptive bibliometric statistics, and statistical information on Web users’ behaviour.

The article is organized as follows. The first two sections deal with the presentation of the Miri@d server (section 2), and the statistical indicators that Miri@d is able to produce (section 3). The results of the Miri@d application are exposed in section 4. Section 5 describes the co-usage analysis and section 6 deals with the application of co-usage analysis on Web user data coming from the Miri@d server.

2 Server organisation

We provide in this section a detailed description of the Miri@d server structure. We start distinguishing the conceptual model that Miri@d represents and its actual technological implementation. The first is general and the second is local.

2.1 The model

Figure 1 represents what we call the model. The model is general in the sense that it is not limited to the particular characteristics represented in figure 2. It is significant to see that the model implies three families of data which it can exploit on the one hand log-files data and on the other hand bibliographic data and commercial data. From the economic point of view, the bibliographic database can be replaced by the concept of an unspecified product database. From the point of view of scientific information, the bibliographic database can also be any. At least theoretically, i.e., on the level of its concept, the model is not completely enclosed within the data sources which it is today using.

Figure 1: The model

2.2 The server structure

Figure 2 represents the server structure, which consists of a set of external resources that providing raw data, and a set of database internal to server.

The resources from which Miri@d receives data:

DM

document delivery management system

CM

customer management system

LM

library management system and

Artic

…(Full text truncated)…