Overview

Monitoring and analyzing multi-variate data from complex systems is a challenge present across various scientific and industrial fields. Traditional methods have problems with real-time anomaly detection and don’t have the ability to show the causal relationships between variables. Our anomaly detection solution (AnomalAIzer) addresses these challenges by combining advanced machine learning algorithms with real-time data processing and causal inference techniques. This solution enables researchers and industries to track, analyze, and respond to complex multi-variate data streams.

Goals

The project goal was to develop a versatile, all-in-one platform capable of tracking and analyzing multi-variate data from various outputs while also providing real-time anomaly detection and causal inference capabilities.

Solution

We deployed AnomalAIzer for real-time anomaly detection in multi-variate data streams. AnomalAIzer consists of two parts, first part compares current data points to previous ones, while the second focuses on changes in between the variables. Use of dual systems allows for great fine-tuning, balancing between detection speed and false alarm prevention. To make the use easier there is also a user-friendly interface that presents findings through clear visualizations, enabling user to quickly identify and respond to detected anomalies.

Results

• Using AnomalAIzer the client was able to prolong post-mortem brain activity. The platform also achieved 99.92% accuracy in identifying significant changes in multi-variate data streams, with an average detection latency of less than 100 milliseconds all while providing real time view of variable relations with it’s causality engine.