Methods and systems for association rule mining
Inventors
Kumar, Anshul • Hookom, Jacob • Tran, Duy
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-11651310-B2
Publication Date
2023-05-16
Expiration Date
Abstract
Provided are methods, systems, and apparatuses for improving supply chain management interfaces and functionality. A frequency of occurrence for each unique combination of a use attribute and a product attribute for various supplies may be determined. A frequency table containing each unique combination may be generated as well as one or more association rules. A level of confidence and a level of lift may be determined for each of the one or more association rules, both of which may be used to generate a score for each unique combination that may be used to rank each unique combination. The ranked unique combinations may then be provided to a supply chain management interface and a database.
Core Innovation
The invention provides a medical data mining method in which a data mining engine generates a frequent pattern tree (FP-tree) comprising a plurality of nodes associated with database records, and each database record comprises a medical diagnosis attribute and a product attribute. The method determines first plurality of links in the FP-tree between nodes that share a common medical diagnosis attribute and a common product attribute, and determines a second plurality of links for each product attribute, where the links are indicative of paths to a root node of the FP-tree.
Using the FP-growth algorithm, the invention determines a frequency of occurrence for each unique combination of medical diagnosis attribute and product attribute, recursively based on the first plurality of links and the second plurality of links in the FP-tree and the corresponding paths to the root node. Based on the frequency of occurrence, it generates a frequency table indicative of each unique combination and a preliminary association rule set in which each association rule comprises a level of confidence and one or more of the product attributes within the plurality of database records.
The invention improves the preliminary association rule set by determining redundancy using confidence, and determines that a first association rule is a redundant association rule when the level of confidence for the first association rule is less than or equal to the level of confidence for a second association rule of the preliminary association rule set. The method removes the redundant association rule to generate a final association rule set, stored in the database, and provides, based on the final association rule set and a query comprising a first medical diagnosis attribute associated with at least one symptom attribute, at least one product identifier and a suggested quantity.
Claims Coverage
The partial content includes three independent claims: clm-00001 (method), clm-00008 (non-transitory computer-readable storage medium), and clm-00015 (apparatus). Across these independent claims, four inventive features are consistently covered.
Frequent pattern tree from diagnosis and product records
Generating, with a data mining engine, a frequent pattern tree (FP-tree) comprising nodes associated with database records, where each database record comprises a medical diagnosis attribute and a product attribute, and the FP-tree is stored in a database associated with the data mining engine; determining links indicative of paths to a root node based on common medical diagnosis attribute and common product attribute.
FP-growth frequency computation for unique diagnosis/product combinations
Determining, by the data mining engine using an FP-growth algorithm, a frequency of occurrence for each unique combination of medical diagnosis attribute and product attribute, recursively based on the first plurality of links and the second plurality of links in the FP-tree and the corresponding paths to the root node; generating a frequency table indicative of each unique combination based on the frequency of occurrence.
Preliminary association rules and redundancy removal by confidence comparison
Determining, based on the frequency table, a preliminary association rule set, each association rule comprising a level of confidence and one or more of the product attributes within the plurality of database records; determining that a first association rule is a redundant association rule based on the level of confidence being less than or equal to the level of confidence for a second association rule; removing the redundant association rule from the preliminary association rule set to generate a final association rule set.
Query-driven provision of product identifiers and symptom-based suggested quantity
Providing, based on the final association rule set, and based on a query comprising a first medical diagnosis attribute associated with at least one symptom attribute, at least one product identifier and a suggested quantity associated with the at least one product identifier, wherein the at least one product identifier comprises at least one product attribute associated with at least one unique combination of the first medical diagnosis attribute and the at least one product attribute, and wherein the suggested quantity is based on the at least one symptom attribute.
Across clm-00001, clm-00008, and clm-00015, the independent claims cover an FP-tree/FP-growth workflow that mines unique medical diagnosis attribute and product attribute combinations, produces association rules with confidence, removes redundant rules using confidence comparisons, and uses a diagnosis/symptom-based query to provide product identifiers and a suggested quantity.
Stated Advantages
Documented Applications
No documented applications found
Interested in licensing this patent?