Which method of machine learning (regression, SVM, NN...) is the best to apply on Neo4j graph database, the aim is just to find the most influencing (important) nodes (parameters, variables) on a specific mesure?

You will likely want to take a look at the Graph Data Science Library. There are several different ways that you can calculate the importance (centrality) of a node, which you can read about here. There are several algorithms there and which one you choose is more related to what your graph is. In order to get a feel for how they differ, you might want to check out the free Graph Algorithms book which walks you through most of the options.

Good luck!