Systems and methods for machine learning based error correction in large data structures

A method comprising:

• storing, by a processor, a plurality health records associated on a many to one basis with each of a plurality of consumer accounts;
• storing, by the processor, a similarity threshold value;
• storing, by the processor, a dissimilarity threshold value;
• storing, by the processor, training data set of health records;
• executing, by the processor, a data processing module configured to apply algorithmic transformations to the health records and comprising a machine learning module including a self learning system, a training system, and a feedback system;
• automatically calculating, by the processor, a first set of weights associated with a first set of factors of a plurality of factors, stored in association with each health record of the training data set of health records;
• automatically entering, by the processor, enters the first set of weights and first set of associated factors into a similarity score model;
• entering, by the processor, at least one factor from the plurality of factors stored in association with each health record associated with the respective consumer account into a similarity score module;
• executing, by the processor, the similarity score module to calculate a similarity score for each one of the plurality of health records associated with the respective customer account, wherein the respective similarity score for the respective health record is based on at least one weight of the first set of weights and at least one factor of the first set of factors of the similarity score model;
• comparing, by the processor, the similarity score for each record with the similarity threshold value and the dissimilarity threshold value to determine a subset of the health records which exceed either value;
• flagging, by the processor, health records as similar or as dissimilar based on the comparing of the similarity score;
• and
• executing, by the processor, a corrective action only for those consumer accounts or records which have associated health records flagged as similar or dissimilar.

An article of manufacture including a non-transitory, tangible computer readable storage medium having instructions stored thereon that, in response to execution by a processor, cause the processor to perform operations comprising:

• storing, by the processor, a plurality health records associated on a many to one basis with each of a plurality of consumer accounts;
• storing, by the processor, a similarity threshold value;
• storing, by the processor, a dissimilarity threshold value;
• storing, by the processor, training data set of health records;
• executing, by the processor, a data processing module configured to apply algorithmic transformations to the health records and comprising a machine learning module including a self learning system, a training system, and a feedback system;
• automatically calculating, by the processor, a first set of weights associated with a first set of factors of a plurality of factors, stored in association with each health record of the training data set of health records;
• automatically entering, by the processor, enters the first set of weights and first set of associated factors into a similarity score model;
• entering, by the processor, at least one factor from the plurality of factors stored in association with each health record associated with the respective consumer account into a similarity score module;
• executing, by the processor, the similarity score module to calculate a similarity score for each one of the plurality of health records associated with the respective customer account, wherein the respective similarity score for the respective health record is based on at least one weight of the first set of weights and at least one factor of the first set of factors of the similarity score model;
• comparing, by the processor, the similarity score for each record with the similarity threshold value and the dissimilarity threshold value to determine a subset of the health records which exceed either value;
• flagging, by the processor, health records as similar or as dissimilar based on the comparing of the similarity score;
• and
• executing, by the processor, a corrective action only for those consumer accounts which have associated health records flagged as similar or dissimilar.