Analyzing Metrics for Tech Mining Methodologies

Extended Abstract – EXTENDING TECHMINING METHODS session at “1st Global TechMining Conference” 2011

Author(s): ARHO SUOMINEN (University of Turku)

What measures or metrics of technological development are visible and practically attainable within bibliometric data? Are the metrics attained objective, reproducible and valid across a variety of technologies?

There is widespread recognition that information embedded in bibliometric databases can be a source of technological foresight data. Databases, such as ISI Web of Science, Compendex, US Patent and Trademark Office, and Espacenet, are sources of rich data when explaining the progression of a specific technology. Approaching databases with different methods e.g. purely quantitative (Järvenpää, Mäkinen, & Seppänen, 2011), cluster analysis (Tseng, Lin, & Lin, 2007), trend extrapolations (Daim, Rueda, Martin, & Gerdsri, 2006), or causal relationships within the data (Kajikawa, Takeda, & Matsushima, 2010) insight on the development of a specific technology has been gathered. Thus creating Competitive Technical Intelligence, as referred to by Porter & Newman (2011), which enables corporations to survive in the ever-dynamic market place. However, discussion on what would be regarded as valid quantitative measures of technological progression and what underlying structures explaining the development can be found within the data are absent from literature.

Motivated by the growing number of bibliometric studies published(1), the focus turns into understanding the quantitative measures that could be valid in analyzing different technologies, thus creating practical tools for managers evaluating different technologies. As such, this study analyses the following research questions: What measures or metrics of technological development are visible and practically attainable within bibliometric data? Are the metrics attained objective, reproducible and valid across a variety of technologies?

To make the questions more approachable, it is seen as useful to explain what is meant by measure or metric in the context of the study. Taking the analogy from Software Metrics, which tries to achieve objective, reproducible and quantitative measures to software development, this study focused on suggesting and evaluating several measures (or metrics) for analyzing bibliometric data. Similarly, to software, tech mining arguable requires objective and reproducible qualitative measures, which are valid with a range of different technologies. Thus, the research uses tech metrics to define measures used to analyze technological development through bibliometric data in an objective and reproducible manner.

The academic literature has not focused on tech metrics as is, but there is considerable prior work from several authors that has created a significant body of knowledge. Going back to 1989 when Robert U. Ayres (1989) called for better methods of forecasting and planning. Focusing especially on quantitative methods of assessing technological development, Ayres thought that more accurate tools for decision-making, on a macroeconomic and microeconomic scale, were needed. Building on previous work on long-range planning and empirical measures of technological development, Ayers sought after decision-making tools that would complement previously used qualitative methods of technological forecasting. Since the number of quantitative studies focusing on different technologies have been published (Borgman and Furner 2002, Kajikawa, Yoshikawa, et al. 2008, Huang, Li and Li 2009, Tseng, Lin and Lin 2007, T. U. Daim, G. Rueda, et al. 2006, Kostoff, et al. 2001). Finally, Porter and Newman (2011) connected the results of quantitative analysis to corporate competitiveness. In the abundance of different quantitative tech mining studies published, interesting case specific results are often produced. Tech mining metrics, which would consistent between technologies, are however lacking.

The analysis in this paper draws from an analysis of two technologies, dye-sensitized solar cells and direct methanol fuel cells. The selected technologies were used to study different approaches of modeling quantitative data and selecting tech mining metrics. As a data source for the study, the ISI web of Science database often seen in bibliometric studies, was used. Through analyzing the possibilities given by the data source, the study went on to suggest several practically attainable metrics that could answer the research question. The metrics created, where then analyzed to validate the measures as objective, reproducible and being valid across different technologies.

The study found, that externally consistent metrics of measuring trajectories are challenging. Although similarities within results between technologies are apparent, the case technologies produced significantly different results. This suggests that while case sensitive trajectory is easily created, more general metrics might even be impractical. However, the low number of technologies studied and that only one database was used limited the study. Thus, further studies on creating consistent metrics for tech mining should be done.

Login to view.

Leave a Reply

Your email address will not be published.