The goal of this master thesis is to analyze the structure of the commercial activities in the cities of Geneva and Barcelona. This study implemented a network‐based approach to predict the location of commercial activities, on the basis of which a tentative inference of urban centrality indices was carried out. This was done by using a ‘grouping algorithm’ and by calculating a ‘location quality index’, both presented and described in the ‘Network‐based predictions of retail store commercial categories and optimal locations’ paper published by Pablo Jensen (2006). In the first part of the work, an existing prototype of the algorithm – implemented during a semester project as VBscript in the GIS software ‘Manifold’ – was optimized. Time gains of a factor 10 for the ‘link computation’ and more stable results for the ‘grouping algorithm’ could be achieved. A spatial representation of the grouping‐results was elaborated and permitted to recognize the dominance areas of the activity‐groups for both cities. The maps created with the ‘location quality index’ (Q‐Index) allowed to identify preference areas and location patterns for the different categories of commercial activities. It was possible to notice that the initial definition or choice of the categories is very important to get clear results. Indeed, the results on Geneva with only 18 categories performed less well than the ones with 45 and 48 categories on Geneva and Barcelona respectively. Furthermore, tests were carried out on the basis of hectometric referenced data, like the database of the Swiss federal census of enterprises. The results obtained are very similar to the ones from the postal address‐based data (precision 2‐3m). This shows that the algorithm used here could be applied to whole Switzerland in a possible future study to determine the overall structure of commercial activities in the country, to identify possible central locations, or also to propose it as useful tool for cantonal or regional economic development agencies to determine optimal locations for new coming large foreign industrial commercial groups. The results of the comparisons between the Q‐Index and centrality indices showed that an approximation of the latter by the first one is not possible. But the Q‐Index (describing location preference per category of commercial activity) can be used as a complementary indication to centrality (which describes accessibility in a broad sense). As a perspective, it would be particularly interesting to work on a definition of a Q‐Index for groups of categories or even of a global integrated one: it would then be probably possible to use this index as a surrogate for the evaluation of value landed property in particular.