A partnership between Eli Z Group and Semeion

Reveal Hidden Relationships in Complex Data

AutoCM is an unsupervised artificial neural network that identifies non-linear relationships traditional statistical methods miss. Transform CSV data into actionable insights with MST, MRG, and H₀ function analysis.

Request API AccessView API DocsReal-world examplesLearn More

Advanced Graph Analysis Methods

AutoCM combines multiple sophisticated algorithms to provide comprehensive insights into your data relationships.

Minimum Spanning Tree (MST)
Reveals the fundamental connections in your dataset by minimizing total energy while maintaining all critical relationships between variables.
Maximally Regular Graph (MRG)
Extends MST by adding relevant cyclic connections, revealing both structure and functional relationships with time-dimension insights.
H₀ Complexity Function
Measures the topological complexity of your data graph, providing a quantitative index from 0 (chain) to 2 (fully connected).
Spectral MRG Clustering
Automatically cluster your graph to identify similar groups, enabling faster analysis and pattern recognition in complex datasets.

Simple API, Powerful Results

Get started with AutoCM in three simple steps. Works with sparse data and small datasets.

01

Upload Your Data

Send a CSV-style table with n-variables (columns) and m-rows of values. Variables can be of different types, enabling cross-domain analysis.

02

AutoCM Processing

Our neural network analyzes relationships using the H₀ function, generating MST and MRG graphs that reveal hidden patterns and connections.

03

Visualize Insights

Receive interactive graphs showing node relationships and their relevance weights. Spectral clustering helps identify similar groups instantly.

Unlock Patterns Across Industries

From retail to healthcare, AutoCM reveals relationships that traditional methods cannot detect.

Product Association

Discover which products are naturally purchased together, even without transaction history. Perfect for merchandising and cross-selling strategies.

Product bundlesInventory optimizationStore layout planning
Geographic Relationships

Identify unexpected connections between products and locations, or customers and regions to optimize distribution and marketing.

Market expansionRegional preferencesSupply chain optimization
Customer Insights

Understand hidden patterns in customer behavior and preferences to suggest new destinations, products, or services they will love.

PersonalizationRecommendation enginesChurn prediction
Multi-Type Analysis

Analyze relationships across different variable types simultaneously: product-country, customer-preference, color-size, and more.

Cross-domain insightsComplex pattern detectionMulti-factor analysis
Real-World Examples

See AutoCM in Action

Each example comes with a description of the dataset, the key insights AutoCM revealed, and a downloadable test dataset so you can try it yourself with the API.

MRG graph of 50 terrorist attacks in Afghanistan showing tribal clusters: Durrani, Jaj, Ghilzai, Safi, Panjsheri, Jadran
MRG + H₀ Function

Security Intelligence & Threat Analysis

Terrorist Attack Patterns in Afghanistan

AutoCM was applied to a dataset of 50 major terrorist attacks against Allied Forces in Afghanistan (up to May 2009). Each attack was characterized by the attacking tribe, ethnic group, and location. A second dataset provided military force estimates per tribe.

Key Insights Revealed

  • 76% of attacks traced to just 4 tribes: Durrani, Jaj, Ghilzai and Safi
  • Geographic clusters mirrored on the Afghanistan map — without using any latitude/longitude data
  • Strong strategic link detected between Durrani and Jaj tribes
  • Safi tribe identified as operating with an independent attack strategy
  • Jadran tribe attacks attributed to strong Ghilzai influence via hidden connections
  • Algorithm split graph into "high-threat" and "low-threat" tribal regions automatically

Input Variables (CSV columns)

TribeEthnic GroupAttack TypeMilitary StrengthArmamentsAttack Frequency
Download Dataset (attacks.zip)Try with the API

Request API Access

Interested in using AutoCM for your data analysis? Fill out the form below and we'll get in touch with you shortly.

Get Started with AutoCM
Share your information and we'll reach out with API access details.