This work is developed over the question ”How to automatically create a good clusteringon spatial dataset with hight different local densities?” opened by previus work of Berzi. To answer the main question, this work describe a approach of recursive clustering pro-cess based on a technique of finding ”vague-solution”, where the output is an hierarchicalclustering of initial dataset. In particularly the the approach is developed and tested on DBSCAN algorithm with large dataset gathered by Google Place in Metropolitan Areaof Milan. The core solutions developed in this algorithm are condensed in the capacity of gener-ation a Hierarchical Clustering with a recursive select the best solutions in according tothe our goals, previously dfined by some sets of rules. The algorithm described here, and developed in my Master Thesis, rosolve two problem: - When we use an algorithm of clustering that can create a set of differents clustering,but equally valid and don’t know exctly what we must have as good solution, we areled to ask ourselves: ”What are the better?” This obviously depends by our goals, sonow the question is: ”What are our goals?” - The second problem is condesned in the sentence ”Not all clusters can be found inone-shot clustering process, more often we must reapply the process to some partof datataset”, so there we have a second question that this paper answering: ”Howto create clustering of data with ad-hoc processing for each different part of inputdataset?” These questions are resolved by the approaches namend in this work as: Space ofSolutions, Vague-Solution, Vague-Solution finding Method and finaly Recursive Clustering. All of these approach was drafetd and testes algoruthm in mine Master Thesis titled ”Geospatial data analysis for Urban informatics applications: the case of the Google Placeof the City of Milan”.