Creating behavioral profiles of your customer
summary
Cleora is an open source graph-embedding tool that converts event streams into compact vectors using simple iterative matrix multiplication on a Markov transition matrix, scaling to tens of millions of users and billions of nodes, running on CPU via a Rust binary with Python bindings. It supports hypergraphs, offers star versus click expansion, updates new nodes by neighbor averaging without full retrain, matches or outperforms Pytorch-BigGraph on large graphs, and includes an MD method plus a commercial Cleora+ for faster production use.
AI-written summary · about
chapters (12)
- 0:00Behavioral profiling with graphs
- 1:52Graph embeddings for behavior modeling
- 4:41Handling massive customer graphs
- 6:50Cliora’s embedding algorithm
- 9:43Cliora vs Python Big Graph
- 11:08Optimization for Cliora’s speed
- 12:53Specialized embeddings for tasks
- 13:59Universal behavior profiles with MD
- 17:19Python implementation of Cliora
- 21:56Visualizing embeddings for classification
- 22:37Cliora Plus for production
- 24:26Q&A
related talks
19mCustomer Journey Analytics using Neo4j Graph Platform
42mBeyond Vectors: Evolving GenAI through Transformative Tools and Methods
29mStay or Go? Predicting Customer Churn through Graph Relationships
12mTopology Over Predictions: Using Graph Context to Drive Real-Time Lending Decisions
40mBuilding Machine Learning Models with Clojure and Cortex
49mModern Graph Databases
28mCreating Enterprise Knowledge Graphs from Unstructured Data
01h 41mGraph Intelligence: Enhance Reasoning and Retrieval Using Graph Analytics