Provlepsis4j: Querying Future Graphs in Neo4j

Abstract

Graphs model relationships in social, information, and collaboration networks, where users often need to reason about both the current graph and its possible evolution. However, graph databases such as Neo4j primarily query the present graph, while link-prediction methods typically run as separate workflows that output ranked candidate edges. We present Provlepsis4j, a Neo4j-based system that integrates link prediction with Cypher querying by maintaining both current and predicted graph states. Provlepsis4j offers an administrator view for configuring, running, and evaluating link-prediction models, and a user view for selecting among future timesteps, querying current and predicted graph states with the same Cypher queries, and inspecting answers side by side. In this way, Provlepsis4j makes predicted graph evolution directly queryable and helps users understand how predicted edges affect graph-query results.