# US 12,001,800 B2 · Semantic-aware feature engineering > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,001,800 B2 - Document ID: US12001800B2 - Status: Granted - Kind: B2 - Application number: 16/567,761 - Filed: 2019-09-11 - Published: 2020-03-19 - Granted: 2024-06-04 - Earliest claimed: 2018-09-13 - Assignees: Feedzai—Consultadoria e Inovação Tecnológica, S.A. - Applicants: Feedzai—Consultadoria e Inovação Tecnológica, S.A. - Inventors: Paulo César Gonçalves Marques; Miguel Ramos de Araújo; Bruno Casal Laraña; Nuno Miguel Lourenço Diegues; Pedro Cardoso Lessa e Silva; Pedro Gustavo Santos Rodrigues Bizarro - Web page: https://patentgrep.com/patent/US12001800B2 ## CPC - G06F 40/30: Semantic analysis - G06F 17/18: for evaluating statistical data {, e.g. average values, frequency distributions, probability functions, regression analysis (forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04)} - G06F 18/2155: {characterised by the incorporation of unlabelled data, e.g. multiple instance learning [MIL], semi-supervised techniques using expectation-maximisation [EM] or naïve labelling} - G06N 20/00: Machine learning - G06Q 20/4016: {involving fraud or risk level assessment in transaction processing} - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR ## Abstract In an embodiment, a process for semantic-aware feature engineering includes receiving semantic labels for data fields of training data. Each of the semantic labels is associated with a semantic meaning associated with a corresponding data field. The process includes automatically generating at least one new feature using at least a portion of the semantic labels. ## Claims ### Claim 1 A method comprising: receiving, by a processor, semantic labels for data fields of training data, wherein each of the semantic labels includes a semantic meaning associated with a corresponding data field; training, by the processor, a machine learning model including by automatically generating at least one new feature to improve interpretability of the machine learning model as compared to not using the at least one new feature including by: determining, by the processor, that a received semantic label of the received semantic labels meets a transformation condition specified by a transformation, wherein the transformation condition defines attributes of a label to which a corresponding transformation is appliable; in response to the determination that the received semantic label meets the transformation condition, applying, by the processor, the transformation to the data fields based at least in part on the semantic labels to determine the at least one new feature including by building a combination of tagged fields by adding the received semantic label and associated data field to a group of tagged fields; wherein a tagged field includes a data field and an associated semantic label, and the transformation condition reduces computer processing resources by limiting a transformation scope for the at least one new feature as compared to not considering the transformation condition; and using, by the processor, the at least one new feature to train the machine learning model; providing, by the processor, the trained machine learning model that uses the at least one new feature that improved the interpretability of the machine learning model and consumed the reduced computer processing resources; and determining, by the processor, a profile characterizing behavior of at least one entity including at least one of users and cards, wherein the profile includes at least one transformation configured to automatically generate the at least one new feature. ### Claim 2 (depends on 1) The method of claim 1, further comprising forming a policy for processing data based at least in part on the at least one new feature. ### Claim 3 (depends on 1) The method of claim 1, wherein automatically generating the at least one new feature includes adding the at least one new feature and associated semantic label to a set of features. ### Claim 4 (depends on 3) The method of claim 3, wherein automatically generating the at least one new feature includes applying a subsequent transformation using the set of features in response to a determination that an iteration condition is not met. ### Claim 5 (depends on 4) The method of claim 4, wherein the iteration condition includes at least one of: a threshold number of determined features, threshold number of iterations, and no new data fields created. ### Claim 6 (depends on 1) The method of claim 1, wherein the transformation is domain-specific. ### Claim 7 (depends on 1) The method of claim 1, wherein applying the transformation includes: calculating a value for the at least one new feature; naming the at least one new feature; and determining at least one semantic label and semantic meaning to assign to the at least one new feature. ### Claim 8 (depends on 7) The method of claim 7, wherein calculating the value for the at least one new feature includes applying a function to a received data field associated with a received semantic label. ### Claim 9 (depends on 7) The method of claim 7, wherein the naming of the at least one new feature includes determining a naming convention for the at least one new feature. ### Claim 10 (depends on 1) The method of claim 1, wherein applying the transformation further includes, prior to determining that the received semantic label meets a transformation condition: creating an instance of the transformation. ### Claim 11 (depends on 10) The method of claim 10, wherein the transformation specifies the transformation condition for input semantic labels such that those data fields with conforming semantic labels are used to determine the at least one new feature. ### Claim 12 (depends on 1) The method of claim 1, wherein the semantic labels are specified in a data schema. ### Claim 13 (depends on 1) The method of claim 1, wherein at least one semantic label of the semantic labels is associated with a composite field, the composite field including a plurality of data fields with respective semantic labels. ### Claim 14 (depends on 1) The method of claim 1, automatically generating the at least one new feature includes applying a transformation over a time window to the data fields based at least in part on the semantic labels to determine the at least one new feature. ### Claim 15 (depends on 14) The method of claim 14, wherein a time window is associated with the received semantic labels. ### Claim 16 (depends on 1) The method of claim 1, further comprising training a machine learning model at least in part by using the automatically generated at least one new feature, wherein the machine learning model is adapted to identify fraud. ### Claim 17 A system comprising: a processor configured to: receive semantic labels for data fields of training data, wherein each of the semantic labels includes a semantic meaning associated with a corresponding data field; train a machine learning model including by automatically generating at least one new feature to improve interpretability of the machine learning model as compared to not using the at least one new feature including by: determining that a received semantic label meets a transformation condition specified by a transformation, wherein the transformation condition defines attributes of a label to which a corresponding transformation is appliable; in response to the determination that the received semantic label meets the transformation condition, applying the transformation to the data fields based at least in part on the semantic labels to determine the at least one new feature including by building a combination of tagged fields by adding the received semantic label and associated data field to a group of tagged fields; wherein a tagged field includes a data field and an associated semantic label, and the transformation condition reduces computer processing resources by limiting a transformation scope for the at least one new feature as compared to not considering the transformation condition; and using the at least one new feature to a machine learning model; provide the trained machine learning model that uses the at least one new feature that improved the interpretability of the machine learning model and consumed the reduced computer processing resources; and determine a profile characterizing behavior of at least one entity including at least one of users and cards, wherein the profile includes at least one transformation configured to automatically generate the at least one new feature. ### Claim 18 A computer program product embodied in a non-transitory computer readable storage medium and comprising computer instructions for: receiving semantic labels for data fields of training data, wherein each of the semantic labels includes a semantic meaning associated with a corresponding data field; training a machine learning model including by automatically generating at least one new feature to improve interpretability of the machine learning model as compared to not using the at least one new feature including by: determining that a received semantic label of the received semantic labels meets a transformation condition specified by a transformation, wherein the transformation condition defines attributes of a label to which a corresponding transformation is appliable; in response to the determination that the received semantic label meets the transformation condition, applying the transformation to the data fields based at least in part on the semantic labels to determine the at least one new feature including by building a combination of tagged fields by adding the received semantic label and associated data field to a group of tagged fields; wherein a tagged field includes a data field and an associated semantic label, and the transformation condition reduces computer processing resources by limiting a transformation scope for the at least one new feature as compared to not considering the transformation condition; and using the at least one new feature to train the machine learning model; providing the trained machine learning model that uses the at least one new feature that improved the interpretability of the machine learning model and consumed the reduced computer processing resources; and determining a profile characterizing behavior of at least one entity including at least one of users and cards, wherein the profile includes at least one transformation configured to automatically generate the at least one new feature.