ontologies for AI & ML
Graph Algorithms
(over Relational Data)
Timbr enables users to leverage powerful graph algorithms directly over SQL ontologies mapped to relational data, no data export, ETL, or external graph database required.
With native SQL execution and full reuse of Timbr-defined relationships, users visualize hidden patterns, detect communities, and discover connections.
Key Capabilities
Run natively on existing databases
Execute graph algorithms directly from Azure SQL, Google BigQuery, Amazon Athena, MySQL, Oracle, SAP HANA, and any other SQL-compatible engine.
Queryable from BI, ML and AI tools
Access and visualize graph outputs directly from Power BI, Tableau, Superset, Qlik, and other popular BI tools using standard SQL or semantic metrics.
Your Choice: CPUs or NVIDIA GPUs
Run on CPUs or NVIDIA GPUs. Materialize from any database to a Graph Framework (NVIDIA RAPIDS cuGraph, Spark GraphFrames, Python NetworkX).
Bi-directional graph enrichment
Results can be automatically written back into the semantic layer to enrich concepts, enabling smarter filtering, recommenda-tions, and ML feature engineering.
Supported Graph Algorithms *
| Algorithm | Description | Typical Use Cases |
|---|---|---|
| PageRank | Computes PageRank scores to assess node importance based on incoming links | Influence ranking (e.g., key accounts, central products, citation networks) |
| Betweenness Centrality | Measures how often a node appears on shortest paths between others | Identifying brokers, bottlenecks, or gatekeepers in networks |
| Katz Centrality | Measures influence considering both direct and indirect neighbors | Propagation modeling, influence scoring beyond direct connections |
| Node Classification | Predicts labels for unlabeled nodes using harmonic functions | Fraud detection, churn prediction, category inference |
| Louvain Community Detection | Detects communities by optimizing modularity in graph partitions | Customer segmentation, topic clustering, entity grouping |
| Strongly Connected Components | Identifies groups where every node is reachable from every other node (directed) | Finding closed ecosystems, transaction loops, feedback systems |
| Weakly Connected Components | Identifies groups connected without regard to direction | Mapping isolated groups, initial cluster detection |
| Core Number | Assigns the largest k-core a node belongs to, based on degree | Influence filtering, structural analysis of dense subgraphs |
| Cycle Basis | Returns a minimal set of cycles that form a basis for all cycles | Anomaly detection, process loop analysis, biological networks |
| Simple Cycles | Enumerates all simple (elementary) cycles in a graph | Detecting feedback loops, supply chain validation |
| Recursive Simple Cycles | Recursively finds cycles in a directed graph | Risk propagation, network stability analysis |
| Common Neighbor Centrality | Scores node pairs by the number of common neighbors | Link prediction, recommender systems, user similarity |
| Jaccard Similarity | Computes similarity based on shared neighbors vs total neighbors | Similarity scoring, entity resolution, lead matching |
| Fuzzy Jaccard Similarity | Fuzzy-matching version of Jaccard similarity | Inexact entity linking, noisy data matching |
| Overlap Coefficient | Computes overlap coefficient (intersection over smaller set size) | Contact tracing, identifying overlapping interests or behaviors |
* Looking for a specific algorithm? Additional graph methods may be available upon request.
