US 12,169,851 B2
Inducing actions in consumer entities
Kristopher John Frutschy, Clifton Park, NY (US)
Assigned to COIN MUTUAL FUNDS, LLC., Griswold, CT (US)
Filed by Coin Mutual Funds LLC, Clifton Park, NY (US)
Filed on Jun. 7, 2019, as Appl. No. 16/435,381.
Claims priority of provisional application 62/683,503, filed on Jun. 11, 2018.
Prior Publication US 2019/0378165 A1, Dec. 12, 2019
Int. Cl. G06Q 30/02 (2023.01); G06K 19/06 (2006.01); G06N 3/08 (2023.01); G06Q 30/0211 (2023.01); G06Q 30/0251 (2023.01)
CPC G06Q 30/0254 (2013.01) [G06N 3/08 (2013.01); G06Q 30/0211 (2013.01); G06K 19/06037 (2013.01)] 23 Claims
OG exemplary drawing
 
1. A method performed by a computing system for providing action inducers to target consumer entities, the method comprising:
accessing historical transaction data samples relating to actions involving a plurality of consumer entities and a provider entity, each historical transaction data sample comprising:
(i) an interactive action involving a respective consumer entity of the plurality of consumer entities and the provider entity, and
(ii) a corresponding previously provided action inducer that induced the respective consumer entity to take the interactive action,
wherein the previously provided action inducer is at least one of (a) a coupon, (b) a trial subscription, or (c) a voucher;
generating training data based on the historical transaction data samples, each training data sample comprising:
(i) a feature vector based on at least one historical data sample, the feature vector comprising values derived from:
(a) the interactive action involving the respective consumer entity and the respective provider entity, and
(b) the corresponding previously provided action inducer, and
(ii) an outcome label for the feature vector indicating whether the previously provided action inducer was successful in inducing the respective consumer entity to take the interactive action;
training a classifier using the generated training data that includes samples of the feature vectors and the outcome labels,
wherein the classifier receives the feature vectors as input and outputs a predicted outcome label;
determining whether the training of the classifier causes an error rate associated with the training of the classifier to fail to satisfy an error rate condition;
in response to the error rate failing to satisfy the error rate condition, retraining the classifier on a subset of the training data including the feature vectors labeled with the outcome respective to the previously provided action inducer, to cause an updated error rate associated with the retraining of the classifier to satisfy the error rate condition;
generating a first target feature vector for a target consumer entity representing a first target action inducer for the target consumer entity,
wherein the first target action inducer is at least one of (a) a second coupon, (b) a second trial subscription, or (c) a second voucher;
applying the retrained classifier on the generated first target feature vector to generate a first predicted outcome of success for the first target action inducer indicating that the first target action inducer will induce the target consumer entity to take a target interactive action; and
when the first predicted outcome of success for the first target action inducer satisfies an inducement criterion, providing the first target action inducer to the target consumer entity.