Machine Learning

Machine Learning

ASI Machine Learning (ASIML)

Machine Learning from Interlink Software
Machine Learning with Interlink Software

ASIML is a machine learning solution for managing incoming event streams and automatically creating and maintaining Service Models in ASI

ASIML is a horizontally scalable solution that utilises machine learning techniques to assist in grouping events into scenarios and service models dynamically. By dedicating discrete ASI nodes to work on different machine learning “features”,
ASI is able to machine learn its environment at scale.

Dynamic Service Model

ASIMLe Standard Model Event (SEM)

Automatically correlate events into scenarios dramatically reducing the volume of events the user must interact with, the machine learning Objective defaults to 99.99% event correlation rate.

Inside each event scenario exists one to many events that have been automatically correlated to that event scenario.

ASIMLe is controlled by the standard event management policies defined in the ASIML dashboard. This is used to prepare and test the Model for our event and metric based Machine Learning.

SEM policies help to train the ASIMLe model algorithm on how to formulate its predictions.

Training ASIMLe

New policies can be created for Pass2 processing on the event stream. Which can include other CAF fields, as well as enrichment fields and system generated fields. Label Token used to drive ticketing. ASIMLe dashboard details some high-level statistics on the Objective as well as view statistics about the performance against the Objective. The dashboard provides the administrator with an interface to train the AI by defining “facts” and “policies” to help the AI with its predictions.

ASIMLe SEM Model Features

Each models default “features” are presented in a balanced scorecard. The SEM models balanced scorecard contains the following

Datasets

Internal Operational Knowledge

An ASIMLe whitelist/blacklist maintained by the users to help populate the facts that control ASIMLe Accessible via the standard user interface including the AID and Query Builder Definitions here help with determining the User score metric for each object

Planned Changes, known Incidents or Problems

ITSM integrations are used to pull data on any Objects that have recently had, or currently have incidents, problem or changes active or planned in the next 24-96 hours This gives ASIMLe knowledge on probabilities of events being produced from these source objects and is used to calculate an ITSM score for each object

Service Models

Queries to existing service models also help to learn about related objects that may also be impacted

ASIMLe SEM Model features

Extra event tokens contain enriched data for the balanced scorecard in the ASIMLe model Event tokens can be used to prioritise the events within each scenario
Share by: