US 12,131,267 B2
Automated training, retraining and relearning applied to data analytics
Mirella Reznic, Irvine, CA (US); Greg Bolcer, Yorba Linda, CA (US); and Anthony Gullotta, Carlsbad, CA (US)
Assigned to Bitvore Corp., Los Angeles, CA (US)
Filed by Bitvore Corp., Los Angeles, CA (US)
Filed on Aug. 19, 2020, as Appl. No. 16/997,532.
Prior Publication US 2022/0058500 A1, Feb. 24, 2022
This patent is subject to a terminal disclaimer.
Int. Cl. G06N 5/04 (2023.01); G06N 5/043 (2023.01); G06N 20/00 (2019.01)
CPC G06N 5/043 (2013.01) [G06N 20/00 (2019.01)] 18 Claims
OG exemplary drawing
 
1. A method for data analysis, the method comprising: partitioning, via a processor, a plurality of records into a plurality of tagged sets, wherein the plurality of tagged sets comprises a positive set, a negative set, and a neutral set; generating, via a user interface, a model according to a first portion of the plurality of tagged sets, wherein: the user interface comprises one or more records that are selectively weighted according to one or more associated boost settings or suppress settings, the user interface comprises a control panel with a first button configured to add new records, a second button configured to delete records, and a plurality of equalizer sliders configured to boost or suppress each of a plurality of selected records, and the user interface is implemented as a set of functions and procedures accessed via a programmatic interface that allows for the creation of applications that access the features that apply boost and suppress settings; evaluating, via the processor, an initial fit of the model according to a second portion of the plurality of tagged sets; adjusting, via the user interface, the model according to the initial fit of the model; and evaluating, via the processor, a final fit of the adjusted model according to a third portion of the plurality of tagged sets;
the adjusted model to determine a single combined score that differentiates inclusion versus exclusion on a scale of scores related to a theme, the plurality of equalizer sliders are configured to provide the scale of scores related to the theme by boosting or suppressing each of the plurality of selected records associated with the theme, the single combined score is an amount of records identified according to the tradeoff of precision and recall, precision is based on false positives, recall is based on missing positives, and the single combined score is determined according to a precision score and a recall score, the adjusted model is retained for inclusion based on the adjusted model being classified as a best performing model according to the single combined score.