Oferta Tecnológica

Dual model for early detection of anomalies

Dual model based on machine learning for early detection of anomalies

Areas of knowledge

Engineering and architecture

Technology’s Protection Status

A software registration application was submitted.

What are we looking for?

Companies interested in licensing this technology are sought.

Contact Us

- Office of Transfer of Research Results
  Building of Central Research Services
  Campus de Elviña, s/n
  15071 A Coruña
- otri@udc.es
- +34 981 167 173

Export PDF

Last Update

10/12/2021

Dual model for early detection of anomalies

Dual model based on machine learning for early detection of anomalies

Areas of knowledge

Engineering and architecture

Technology’s Protection Status

A software registration application was submitted.

What are we looking for?

Companies interested in licensing this technology are sought.

Description

Machine learning model designed for early detection of anomalies. The model considers time as a relevant factor when detecting a possible anomaly, so that a late detection, although correct, is considered erroneous. The dual model is based on two independent machine learning models: one for the detection of anomalies and another one for the detection of normal cases. The system will provide three possible outputs for an observation: normal, anomaly or delay. In the case of delay, it will be necessary to provide more information to the model to obtain a definitive output (i.e. normal or anomaly).

Added Value

This model was initially developed for the early detection of depression cases based on posts in social networks, improving the results of the state of the art taking into account in the evaluation both the correction of the detection and the time spent in the detection.

In general, we consider that this method can be of interest for any environment or application where time is considered a decisive factor when detecting potential anomalies. The early detection of an anomaly makes it possible to anticipate and deal more quickly with possible negative repercussions in the system, with the consequent economic savings.

Applications according to Sector

This technology is considered transversal and of potential application in virtually any sector where anomaly detection is sought and the time required for that detection is considered a critical factor.