Skip to main content

Intro

Connhex AI is a modular machine learning system that can be used to process data from Connhex Cloud. Its core strength consists in being tailored to processing time series recorded from connected devices - like measurements, events and so on.

Connhex AI: architecture.

Main features​

Connhex AI focuses on anomaly detection and forecasting: it doesn't make any assumptions on your data, so it can be used out of the box.

Models automatically improve over time thanks to the feedback loop with Connhex Cloud: when enough data is available, the model is automatically retrained and updated if the performance is better than the previous one. See here for additional details.

Anomaly detection​

  • Univariate anomaly detection
  • Multivariate anomaly detection
  • Ensembles with automated anomaly score normalization
  • Adjustable monitoring window size​
  • Customizable multi-event handling​

Time series forecasting​

  • Univariate forecasting
  • Multivariate forecasting
  • Variable prediction period​
  • Ensembles: combine the outputs of multiple models to achieve better performance
  • Exogenous regressors: variables that might have an impact on the future values of the variable to predict
  • Prediction quality continuous monitoring​

Integration with Connhex Cloud​

Time Series format

Connhex AI only supports time series stored according to the Connhex Message Policy.

Connhex AI integrates with Connhex Cloud to generate actions from predictions and anomalies. This is enabled by virtual connectables producing a stream of forecasts or anomalies that can be further interpreted by Connhex Cloud services. The most common use-case is receiving an alert when an anomaly is detected: this can be achieved by creating a rule over the generated anomalies stream.

Connhex AI is also integrated with Connhex Edge to perform on-device inference.