This technology is an anomaly detection solution which provides alerts in real-time, when it recognizes an abnormal pattern. It uses various types of data streams to a model based on deep learning neural network. The technology identifies the anomalous patterns by also leveraging on rule-based methods.
The disadvantages of existing competing technologies are:
• Other rule-based methods requires specific domain experts to develop rule sets and is not robust, but fragile and unable to identify slightly different patterns
• Data-driven machine learning approaches require vectorization of data records, and requires the total data set to be stored if upgrades are carried out after deployment
Technology Features, Specifications and Advantages
• Data Digitization : converting non-digital data fields to digital vectors
• Time-series Modeling : applying a neural network model reflecting time-series characteristics
• Hierarchical Behavioural Knowledge Space (HBKS) : storing the anomalous patterns identified by the administrator and applying them in real time. By storing the causes identified by the administrators, it becomes possible to explain the cause of similar phenomena in the future.
• Combining a deep neural network and a rule-based model
• Not requiring labeling of data for machine learning
• Supporting automatic learning
The technology can be applied to products/application listed below:
• Medical diagnosis devices (abnormal pattern detecting devices)
• Finance Industry
- Fraud Detection Systems (FDS)
• Security Industry
- Confidential documents/files
- Network Intrusion Monitoring, Digital Rights Management (DRM) Risk Monitoring
• Manufacturing Industry
- Predictive maintenance in Smart Factories
• Visualization of the data sets
The company is seeking partners to conclude license and commercial agreements to enter the international market. The company is open to all the types and sizes of partners.