Lead your team forward
OCT 24 / 9AM ET Register nowAI Services
Ataccama ONE Platform includes several AI components used to enhance the metadata and categorize it or detect irregularities in your data that affect its quality. The components are as follows:
-
AI Matching services: AI Matching Manager and Worker
-
Anomaly Detection service: Anomaly Detector
-
Term Suggestions services: Api, Feedback, Recommender, Neighbors
Each of the components listed receives input directly from the UI or operates on metadata. The following sections provide more information about how these groups of services function.
Anomaly Detection
Anomaly detection contributes to fostering relevant, correct, and reliable data that evolves together with your business and its processes.
For instance, in the following data set {1, 1, 2, 1, 1, 1}
, 2
is clearly an anomaly.
However, if we include more values and the set now contains an equal number of ones and twos ({1, 1, 2, 1, 2, 2, 2, 1, 2, 1}
), 2
is no longer considered an anomaly.
Likewise, a value that was considered normal can become anomalous after the data changes.
When an anomaly is detected in the data, users are notified of the issue and prompted to confirm or reject the detected result. Based on this user feedback the algorithm learns and becomes more sensitive and accurate over time. Anomaly detection works on the level of individual catalog items, which means that it does not apply what it has learned from one catalog item to another.
AI Matching
The AI Matching is available only if your version of Ataccama ONE includes ONE Master Data Management (MDM). The feature combines user input with AI capabilities to improve results of rule-based record matching, leading to quicker and more efficient elimination of duplicate records in the data catalog, as well as to create new matching rules for ONE MDM using AI-based rule suggestions. User feedback is provided during model training and when resolving matching proposals.
Term Suggestions
The Term Suggestions components make up the logic behind term suggestions and work together to determine which terms best fit your data. For example, they help you identify data sets with personal information, such as first and last name, telephone number, address, social security number. Based on your input, that is, the feedback that you provide when you accept or reject a suggested term, the algorithm learns from you and finetunes future suggestions.
For example, an attribute (column) that was not associated with a term before can after a while be identified as the customer identifier since you have marked other attributes with very similar data as such.
If a similar attribute appears in your other data sources, the attribute will now be correctly recognized and the Customer ID
term will be assigned to it.
Was this page useful?