| CPC G06F 16/2237 (2019.01) [G06F 16/24575 (2019.01); G06F 16/24578 (2019.01); G06F 21/6245 (2013.01); G16H 20/10 (2018.01)] | 16 Claims |

|
1. A method for maintaining user privacy and anonymizing user data when collecting information from a user of a recommendation system to generate a user-specific recommendation by the recommendation system, the method comprising:
at a computing system, receiving from a client device in communication with the computing system and operated by the user, information comprising:
a set of personal care objective identifiers, each personal care objective identifier corresponding to a respective one personal care objective reported by the user;
a set of demographic identifiers, each demographic identifier corresponding to a respective one demographic attribute reported by the user as describing the user; and
a set of environmental identifiers, each environmental identifier corresponding to a respective one environmental attribute reported by the user and/or reported by the client device as describing a location occupied by the user;
generating, by the computing system, a first hash by providing as input to a one-way hashing function the set of demographic identifiers;
generating, by the computing system, a second hash by providing as input to the one-way hashing function the set of environmental identifiers;
generating, by the computing system, a third hash by providing as input to the one-way hashing function the set of personal care objective identifiers;
providing the first hash, the second hash, and the third hash as input to machine learning model having been trained against a structured dataset comprising:
a first dimension defined by a set of hashes based on demographic or environmental attributes extracted from and/or derived from public customer review data authored by anonymous review authors in respect of one or more consumer products;
a second dimension defined by a set of attributes of consumer products referenced in the public customer review data, the set of attributes extracted from at least one public resource;
a third dimension defined by a set of hashes based on personal care objectives extracted from and/or derived from the public customer review data; and
normalized values corresponding to customer review author sentiment extracted and/or derived from of the public customer review data;
receiving, as output from the predictive model, a set of consumer product attributes that, if exhibited by a product used by the user, would be likely to elicit a positive sentiment review from the user;
instructing to be manufactured a custom product for the user, the custom product exhibiting the set of consumer product attributes received as output from the predictive model;
providing the custom product to the user; and
updating the structured dataset and re-training the machine learning model in response to receiving input from the user regarding the custom product.
|