# US 2021/0241293 A1 · Apparatuses, computer-implemented methods, and computer program products for improved model-based determinations > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2021/0241293 A1 - Document ID: US20210241293A1 - Status: Published application - Kind: A1 - Application number: 17/071,177 - Filed: 2020-10-15 - Published: 2021-08-05 - Earliest claimed: 2014-02-28 - Assignees: Groupon, Inc. - Applicants: Groupon, Inc. - Inventors: Nikhil Ranjan GARGE - Web page: https://patentgrep.com/patent/US20210241293A1 ## CPC - G06Q 30/0202: Market predictions or forecasting for commercial activities - 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 Methods, apparatuses, and computer program products are provided that utilize specially configured model-based trees to output data that more accurately satisfies a particular predictive task. Each model-based tree may include any number of branches partitioning an original data set into subsets for processing. Each branch may end in one or more leaf nodes, where each leaf node is associated with a data model specially configured based on a particular set of partitioned data associated with the leaf node. In this regard, the plurality of model-based trees may generate data utilized to determine and/or output preferred data for a particular user. ## Claims ### Claim at position 1 1-57. (canceled) ### Claim 58 An apparatus comprising at least one processor and at least one memory having computer-coded instructions stored thereon that, in execution with the at least one processor, configures the apparatus to: receive an original data set corresponding to a plurality of predictor parameters; generate, based at least in part on the original data set, a plurality of model-based decision trees, each model-based decision tree comprising one or more branches, wherein each branch of the one or more branches is associated with a data model configured based at least in part on a partitioned subset of the original data set determined based on at least one instability parameter corresponding to each decision point in the model branch; apply a plurality of percentage differential data values to each decision tree of the plurality of predictor parameters to generate a set of model output data; aggregate the set of model output data to generate preferred output data associated with the plurality of percentage differential data values and the set of model output data; and output a representation of the preferred output data to a computing device. ### Claim 59 (depends on 58) The apparatus according to claim 58, wherein to generate the plurality of model-based decision trees, the apparatus is configured to: recursively perform: in an instance where a decision point of the one or more decision points is determined not to comprise a leaf node: determine, for a decision point of the one or more decision points, an instability parameter from one or more predictor parameters of the plurality of the predictor parameters; identify at least a first subset of a source data set of the original data set based on the instability parameter, the first subset of the source data set corresponding to a first sub-branch from the decision point, and a second subset of the source data set of the original data set based on the instability parameter, the second subset of the source data set corresponding to a second sub-branch from the decision point; and in an instance where a decision point of the one or more decision points is determined to comprise the leaf node: generate the data model corresponding to the leaf node based on the source data set. ### Claim 60 (depends on 58) The apparatus according to claim 58, wherein the plurality of model-based decision trees each comprise at least one decision point associated with splitting at least a portion of the original data set based on a random subset of the plurality of predictor parameters. ### Claim 61 (depends on 58) The apparatus according to claim 58, wherein the plurality of model-based decision trees each comprise at least one decision point associated with splitting at least a portion of the original data set based on a highest instability parameter of the plurality of predictor parameters. ### Claim 62 (depends on 58) The apparatus according to claim 58, wherein to output the representation of the preferred output data to the computing device, the apparatus is configured to: cause rendering, to the computing device, of a user interface comprising at least the preferred output data. ### Claim 63 (depends on 58) The apparatus according to claim 58, wherein the representation of the preferred output data comprises a preferred percentage differential data value of the plurality of preferred percentage differential data value. ### Claim 64 (depends on 58) The apparatus according to claim 58, wherein to aggregate the model output data, the apparatus is configured to: generate a plurality of aggregated output data values corresponding to the plurality of percentage differential data values by: for each particular percentage differential data value of the plurality of percentage differential data values, aggregating a subset of model output data from the set of model output data, the subset of model output data comprising output from each model-based decision tree of the plurality of model-based decision trees for the particular percentage differential data value of the plurality of percentage differential data values to generate an aggregated output data value corresponding to the particular percentage differential data value; combine the plurality of aggregated output data values to generate the preferred output data; and output a preferred percentage differential data value of the plurality of preferred percentage differential data value and one or more of the plurality of aggregated output data values. ### Claim 65 A computer-implemented method comprising: receiving an original data set corresponding to a plurality of predictor parameters; generating, based at least in part on the original data set, a plurality of model-based decision trees, each model-based decision tree comprising one or more branches, wherein each branch of the one or more branches is associated with a data model configured based at least in part on a partitioned subset of the original data set determined based on at least one instability parameter corresponding to each decision point in the model branch; applying a plurality of percentage differential data values to each decision tree of the plurality of predictor parameters to generate a set of model output data; aggregating the set of model output data to generate preferred output data associated with the plurality of percentage differential data values and the set of model output data; and outputting a representation of the preferred output data to a computing device. ### Claim 66 (depends on 65) The computer-implemented method according to claim 65, wherein generating the plurality of model-based decision trees comprises: recursively performing: in an instance where a decision point of the one or more decision points is determined not to comprise a leaf node: determining, for a decision point of the one or more decision points, an instability parameter from one or more predictor parameters of the plurality of the predictor parameters; identifying at least a first subset of a source data set of the original data set based on the instability parameter, the first subset of the source data set corresponding to a first sub-branch from the decision point, and a second subset of the source data set of the original data set based on the instability parameter, the second subset of the source data set corresponding to a second sub-branch from the decision point; and in an instance where a decision point of the one or more decision points is determined to comprise the leaf node: generating the data model corresponding to the leaf node based on the source data set. ### Claim 67 (depends on 65) The computer-implemented method according to claim 65, wherein the plurality of model-based decision trees each comprise at least one decision point associated with splitting at least a portion of the original data set based on a random subset of the plurality of predictor parameters. ### Claim 68 (depends on 65) The computer-implemented method according to claim 65, wherein the plurality of model-based decision trees each comprise at least one decision point associated with splitting at least a portion of the original data set based on a highest instability parameter of the plurality of predictor parameters. ### Claim 69 (depends on 65) The computer-implemented method according to claim 65, wherein outputting the representation of the preferred output data to the computing device comprises: causing rendering, to the computing device, of a user interface comprising at least the preferred output data. ### Claim 70 (depends on 65) The computer-implemented method according to claim 65, wherein the representation of the preferred output data comprises a preferred percentage differential data value of the plurality of preferred percentage differential data value. ### Claim 71 (depends on 65) The computer-implemented method according to claim 65, wherein aggregating the model output data comprises: generating a plurality of aggregated output data values corresponding to the plurality of percentage differential data values by: for each particular percentage differential data value of the plurality of percentage differential data values, aggregating a subset of model output data from the set of model output data, the subset of model output data comprising output from each model-based decision tree of the plurality of model-based decision trees for the particular percentage differential data value of the plurality of percentage differential data values to generate an aggregated output data value corresponding to the particular percentage differential data value; and combining the plurality of aggregated output data values to generate the preferred output data. ### Claim 72 A computer program product comprising a non-transitory computer readable medium having computer program instructions stored therein that, in execution with at least one processor, are configured for: receiving an original data set corresponding to a plurality of predictor parameters; generating, based at least in part on the original data set, a plurality of model-based decision trees, each model-based decision tree comprising one or more branches, wherein each branch of the one or more branches is associated with a data model configured based at least in part on a partitioned subset of the original data set determined based on at least one instability parameter corresponding to each decision point in the model branch; applying a plurality of percentage differential data values to each decision tree of the plurality of predictor parameters to generate a set of model output data; aggregating the set of model output data to generate preferred output data associated with the plurality of percentage differential data values and the set of model output data; and outputting a representation of the preferred output data to a computing device. ### Claim 73 (depends on 72) The computer program product according to claim 72, wherein to generate the plurality of model-based decision trees, the computer program product is configured for: recursively performing: in an instance where a decision point of the one or more decision points is determined not to comprise a leaf node: determining, for a decision point of the one or more decision points, an instability parameter from one or more predictor parameters of the plurality of the predictor parameters; identifying at least a first subset of a source data set of the original data set based on the instability parameter, the first subset of the source data set corresponding to a first sub-branch from the decision point, and a second subset of the source data set of the original data set based on the instability parameter, the second subset of the source data set corresponding to a second sub-branch from the decision point; and in an instance where a decision point of the one or more decision points is determined to comprise the leaf node: generating the data model corresponding to the leaf node based on the source data set. ### Claim 74 (depends on 72) The computer program product according to claim 72, wherein the plurality of model-based decision trees each comprise at least one decision point associated with splitting at least a portion of the original data set based on a random subset of the plurality of predictor parameters. ### Claim 75 (depends on 72) The computer program product according to claim 72, wherein the plurality of model-based decision trees each comprise at least one decision point associated with splitting at least a portion of the original data set based on a highest instability parameter of the plurality of predictor parameters. ### Claim 76 (depends on 72) The computer program product according to claim 72, wherein to output the representation of the preferred output data to the computing device, the computer program product is configured for: causing rendering, to the computing device, of a user interface comprising at least the preferred output data. ### Claim 77 (depends on 72) The computer program product according to claim 72, wherein to aggregate the model output data, the computer program product is configured for: generating a plurality of aggregated output data values corresponding to the plurality of percentage differential data values by: for each particular percentage differential data value of the plurality of percentage differential data values, aggregating a subset of model output data from the set of model output data, the subset of model output data comprising output from each model-based decision tree of the plurality of model-based decision trees for the particular percentage differential data value of the plurality of percentage differential data values to generate an aggregated output data value corresponding to the particular percentage differential data value; and combining the plurality of aggregated output data values to generate the preferred output data.