Lynx Analytics is open-sourcing LynxKite, its Complete Graph Data Science Platform

Check out this article for a brief summary on what LynxKite is, where it is coming from and how it can help with your data science projects.

By András Németh, CTO of Lynx Analytics

Singapore based Lynx Analytics is a pioneer in driving business value from graph based insights. Our international clientele range from large telcos through banks to transport agencies and governments. Our projects vary from graph based advanced data mapping, demography estimation, customer satisfaction monitoring, fraud detection, fiber network and ATM location optimization, IoT analytics. The common ingredient in all these projects is graph analytics: we always beat state of the art by somehow using the extra information available in relationship data.

One of the important sources of Lynx's success is LynxKite. The company soon realized that the tooling for efficient, interactive, productizable graph mining is very sparse and we created a dedicated R&D team in 2014 in Budapest to build such a platform from scratch. LynxKite was developed hand in hand with our client project teams. We always strived to make our data scientists more efficient, our graph models more powerful, production operations smoother and, ultimately, our clients happier.

Six years and almost 16000 commits later, today we are extremely proud and excited to announce putting this tool open source in the hands of the broader community. We are confident that LynxKite can be an important tool in the hands of data scientists around the world and it will help boost the adoption of often neglected graph methods.

So what is LynxKite exactly and how can it help you?

LynxKite is a graph data science platform. It is to network data what e.g., RapidMiner, SPSS Modeler or Knime is to tabular data.

LynxKite isn't yet-another graph database. It is not about storing and serving graph data, it is about mining insights and building models on top of it. It can be complementary to an existing graph database deployment or it can turn information in traditional data sources into graphs.

With LynxKite you can:

  • Import data (up to terabytes) from a variety of sources. Work directly with traditional data sources (CSV, JSON, ORC, Parquet files — local or Hadoop; JDBC, Hive, etc.) or from a graph DB like Neo4j.
  • Turn data easily into graphs.
  • Use algorithms from a large library of graph operations, including graph neural network operations.
  • Put together complex data processing pipelines, where you can combine graph operations, classical data analysis operations and machine learning.
  • Discover graphs and interpret algorithm results interactively, at any stage or step of the calculations, easily experimenting with different approaches and tuning parameters.
  • Seamlessly combine the benefits of a friendly GUI and optional coding via a powerful Python integration (code embedding, Python API, code generation).
  • Accelerate adoption in your organization by creating your own wizards that allow less technical people to use your sophisticated graph models.

LynxKite is an abstract, generic tool, good for anything that involves analyzing graphs. But this makes defining what it does, as demonstrated above, well, abstract. So, to put a bit of life into the above points, let us finish with a few sneak peeks of LynxKite in action!

Training and applying graph neural networks with just a few click



Click here to see an example workspace!

Using age-homogeneous communities to estimate age



See a complete tutorial for this use case here.

Figuring out what connects Superman to Gandalf (in the knowledge graph)...



... and you can check out other pairs of entities using this wizard!

The list goes on, but this article has to finish... So, check our documentation for all the gory details, reach out to us on with any questions or feedback, and, most of all, happy graph analytics with LynxKite to you all!

Bio: András Németh is a Software Engineer and Mathematician, and the CTO of Lynx Analytics.