DiagFit version 3.4 is here! Discover this new version dedicated to user experience and models performances.
The customer experience and consumer pathway are enhanced by
- A redesigned help path: new help elements have been added to support users at every stage of the path, helping them to make the right choices.
- A better description of models: a new display presents the different models available on DiagFit. It explains how they work, their advantages and disadvantages, and how they complement each other.
- A file size indicator: new indications have been introduced to give the user a better understanding of the amount of data stored in the database.
Model performance has also been optimized by:
Improved data processing: faster, more efficient analysis of import data, improved calculation and training times.
A new “accepted FA” (False Alarm) parameter: this allows the user to specify the acceptable level of false positives in the supposedly healthy training zone during voting, offering greater accuracy in the detection threshold.
A new “window step” parameter: This new feature lets you specify the desired time interval between 2 predictions, and thus control the frequency of alarms in the event of anomaly detection.
Finally, data preparation functions have been enhanced:
New mathematical operators for virtual sensors: calculation of the min or max between two signals, calculation of absolute values, enabling new time series to be added as model inputs.
Synchronization of multi-sensor signals: to enable the generation of multivariate models and thus obtain adequate comparative analyses, it is now possible to synchronize all sensors at a given frequency.
Functionality updates: parameters for the “Nan impute” have been added, for more flexible management of missing data completion. Additional options for visualizing spectra and spectrograms to improve analysis resolution. Global outlier analysis added.
To know more about DiagFit‘s new features:
