| dc.description.abstract | This research presents a network analysis of knowledge production. We analyse a 2-mode network of cities and topics, obtained by a bibliometric case study of publication record of the scientific field of transportation. The main objective is to explore knowledge diffusion by path- and place-dependent patterns. Based on cognitive and geographic proximity of cities and topics we design prediction models to measure this exploratory efforts. The topics we study are based on the title words of articles, written by scientists that work in a certain city while publishing their knowledge. The 2-mode network we study has been transformed in both a city and topic network for further analysis. In these networks we explore the concepts of geographic and cognitive proximity. We linked the optimisation of proximities to the notion of absorptive capacity, which brings the network of cities and topics together. Although this link has been explored before, no quantitative support was found so far. This research shows a high significance level of chi-squared statistics for path- and place-dependency. Since a prediction should be as precise and specific as possible, we also evaluate the prediction models with the more intuitive F-score, which is determined by a weighted combination of a prediction's precision and recall. The precision of our path-dependent prediction model is high for small prediction sets, while the place-dependent model is not very precise. The recall is for both models relatively low. When we optimise the topic proximity with a co-evolutionary approach of the behaviour of cities and topics, the recall of the path-dependent model increases. However, this is at the cost of its precision. A concrete preference for the relative importance of precision and recall is required to determine an optimal design of the prediction model. This design can be either path-dependent, place-dependent or based on a co-evolution of cities and topics. | |
