Evaluating criticality classifications of items using binary classification models and a validation of similar items

An apparatus comprising:

• at least one processing device comprising a processor coupled to a memory;
• the at least one processing device being configured to implement the following steps:
• accessing, using the at least one processing device, one or more data structures comprising information characterizing at least one item having at least one designated criticality classification;
• applying at least a portion of the information characterizing the at least one item to an encoder model that generates an encoded vector characterizing the at least one item;
• applying the encoded vector characterizing the at least one item to at least one processor-based binary classification model, wherein the processor-based binary classification model identifies one or more candidate false positive items, having the at least one designated criticality classification, as candidate false positive classifications;
• evaluating at least one of the one or more candidate false positive items, by performing the following steps, using the at least one processing device, wherein the evaluating is automatically performed by the at least one processing device responsive to an occurrence of at least one of the candidate false positive classifications:
• obtaining a plurality of encoded vectors characterizing respective ones of a plurality of previously processed items having the at least one designated criticality classification;
• comparing the encoded vector characterizing the at least one candidate false positive item to the encoded vectors characterizing the respective ones of the plurality of previously processed items, having the at least one designated criticality classification, to obtain respective similarity scores;
• identifying at least one of the encoded vectors characterizing the respective ones of the plurality of previously processed items having a similarity score, with respect to the encoded vector characterizing the at least one candidate false positive item, that satisfies one or more similarity criteria;
• and
• determining whether a classification of a given previously processed item, associated with the at least one encoded vector having the similarity score that satisfies the one or more similarity criteria, was validated, by a designated validation process, with a false positive classification;
• and
• initiating at least one automated action based at least in part on a result of the evaluating, wherein the at least one automated action comprises one or more of:
• classifying the at least one candidate false positive item as a false positive classification in response to the determining indicating that the classification of the given previously processed item, associated with the at least one encoded vector having the similarity score that satisfies the one or more similarity criteria, was validated with the false positive classification;
• classifying the at least one candidate false positive item with an indeterminate classification in response to the determining indicating that the classification of the given previously processed item, associated with the at least one encoded vector having the similarity score that satisfies the one or more similarity criteria, was not validated with the false positive classification;
• and
• assigning at least one of one or more candidate false positive items to a set of items requiring a manual review.

A non-transitory processor-readable storage medium having stored therein program code of one or more software programs, wherein the program code when executed by at least one processing device causes the at least one processing device to perform the following steps:

• accessing, using the at least one processing device, one or more data structures comprising information characterizing at least one item having at least one designated criticality classification;
• applying at least a portion of the information characterizing the at least one item to an encoder model that generates an encoded vector characterizing the at least one item;
• applying the encoded vector characterizing the at least one item to at least one processor-based binary classification model, wherein the processor-based binary classification model identifies one or more candidate false positive items, having the at least one designated criticality classification, as candidate false positive classifications;
• evaluating at least one of the one or more candidate false positive items, by performing the following steps, using the at least one processing device, wherein the evaluating is automatically performed by the at least one processing device responsive to an occurrence of at least one of the candidate false positive classifications:
• obtaining a plurality of encoded vectors characterizing respective ones of a plurality of previously processed items having the at least one designated criticality classification;
• comparing the encoded vector characterizing the at least one candidate false positive item to the encoded vectors characterizing the respective ones of the plurality of previously processed items, having the at least one designated criticality classification, to obtain respective similarity scores;
• identifying at least one of the encoded vectors characterizing the respective ones of the plurality of previously processed items having a similarity score, with respect to the encoded vector characterizing the at least one candidate false positive item, that satisfies one or more similarity criteria;
• and
• determining whether a classification of a given previously processed item, associated with the at least one encoded vector having the similarity score that satisfies the one or more similarity criteria, was validated, by a designated validation process, with a false positive classification;
• and
• initiating at least one automated action based at least in part on a result of the evaluating, wherein the at least one automated action comprises one or more of:
• classifying the at least one candidate false positive item as a false positive classification in response to the determining indicating that the classification of the given previously processed item, associated with the at least one encoded vector having the similarity score that satisfies the one or more similarity criteria, was validated with the false positive classification;
• classifying the at least one candidate false positive item with an indeterminate classification in response to the determining indicating that the classification of the given previously processed item, associated with the at least one encoded vector having the similarity score that satisfies the one or more similarity criteria, was not validated with the false positive classification;
• and
• assigning at least one of one or more candidate false positive items to a set of items requiring a manual review.