The current literature on party-group relationships finds that some strongly institutionalized relationships still exist. This paper builds upon this and seeks to identify different types of party-group relationships by employing clustering analysis on a unique dyadic dataset generated from survey data. The dyadic dataset covers groups' relationship to parties in Denmark, France, Germany, Netherlands, Norway, UK and US. What are the most frequent types of party-group relationships in these countries? Are there substantial country level differences? Can we talk about different networks of allies? How common are the strongly institutionalized relationships in the different countries/political systems? By utilizing the machine learning technique k-means clustering, the analysis attempts to answer these questions. Clustering divides the data into homogenous groups and portrays the landscape of party-group relationships. The preliminary analysis shows that the largest group of dyads can be labelled as having distant or non-existing relationships. Still, the number of dyads that could be labelled as "frequent collaborators" is substantial. This group of party-group dyads can be described as interacting strategically for specific purposes. The analysis also finds a small number of tightly interlinked party-group relationships. These shorthand summaries of a complicated data structure are further supported by case specific examples. Furthermore, the typologies created may be used to predict party-group behaviour.
