US 12,271,827 B2
Computer prediction of relevant data from multiple disparate sources
Yair Horesh, Kfar Sava (IL); Alexander Zhicharevich, Petah Tikva (IL); Shlomi Medalion, Lod (IL); and Natalie Bar Eliyahu, Tel Aviv (IL)
Assigned to Intuit Inc., Mountain View, CA (US)
Filed by Intuit Inc., Mountain View, CA (US)
Filed on Oct. 30, 2020, as Appl. No. 17/086,317.
Prior Publication US 2022/0138592 A1, May 5, 2022
Int. Cl. G06N 5/022 (2023.01); G06F 18/214 (2023.01); G06F 40/30 (2020.01); G06N 5/045 (2023.01); G06N 20/00 (2019.01)
CPC G06N 5/022 (2013.01) [G06F 18/214 (2023.01); G06F 40/30 (2020.01); G06N 5/045 (2013.01); G06N 20/00 (2019.01)] 17 Claims
OG exemplary drawing
 
1. A method comprising:
extracting data from a plurality of disparate data sources, wherein the data comprises data pairs comprising a corresponding data point and a corresponding time associated with the corresponding data point;
extracting a plurality of insights from the data at least by identifying a trend in the data pairs;
forming a model vector comprising the plurality of insights and an additional attribute to the plurality of insights, wherein the additional attribute characterizes the insights, wherein the additional attribute includes at least user feedback comprising a user ranking of a ranked subset of the insights from a user, and wherein forming the model vector comprises:
converting the plurality of insights and the additional attribute into a data structure comprising a one dimensional matrix formatted and stored for input to a trained insight machine learning model, wherein converting performs a data structure transformation from the plurality of insights and the additional attribute to the one dimensional matrix, and wherein the model vector comprises the one dimensional matrix;
inputting the model vector into the trained insight machine learning model, wherein the trained insight machine learning model comprises a supervised machine learning model configured for regression analysis;
executing the trained insight machine learning model on the model vector to obtain, as output, a predicted ranking of the plurality of insights;
selecting, based on the predicted ranking, a pre-determined number of insights to form predicted relevant insights;
creating a corresponding contextual description for each of the predicted relevant insights by inserting one or more words from the predicted relevant insights into a sentence template; and
publishing a text report comprising the corresponding contextual description and a link associated with the corresponding contextual description, wherein the link is selectable by the user to display at least a portion of the data.