Optimization of Retrieval Algorithms on Large Scale Knowledge Graphs

An efficient approach towards the generation and analysis of interoperable clinical data in a knowledge graph

“…( numberOfPatients DESC LIMIT 3 Since the incident edges of each node are counted, for a dense graph G = (V, E) we obtain a complexity of O(|V | 2 ) in the worst case, i.e. every node v ∈ V is incident to every other u ∈ V, u = v. In [22], it is already shown that the algorithms implemented in Neo4j have such a high running time on large networks that they become practically almost useless. However, the runtimes and the average of query Q9 shown in Fig.…”

“…These are ordered and categorized within this section in order to later analyze their efficiency in VI and to consider comparative values by means of complexity theory. Preliminary work has been carried out in [21] and [22]. There, biomedical questions are examined and optimized as well.…”

“…The simplest query in the class of Graph Navigation Queries is whether a certain path exists in the graph, see [22]. Path queries specified with regular expressions are commonly referred to as Regular Path Queries (RPQ), see [23].…”

“…As discussed in [22], queries become slower with increasing number of queried attributes and used relations. Here, we first consider the queried attributes.…”

“…Conjunctive queries and RPQs can be combined in the class CRPQ, see [23], which then can be extended further by the extended CRPQs which include the possibility to specify path variables and even allows the output of a query to be a path. According to the schema presented in [22] they both are subproblems of pattern matching in graphs. We consider both CRPQs (Q4, Q6, Q8) and ECRPQ (Q10, see Fig.…”