Deep cognitive constrained filtering for product recommendation

A computing system, comprising:

• a processor;
• a memory device coupled to the processor;
• and
• a computer readable storage device coupled to the processor, wherein the storage device contains program code executable by the processor via the memory device to implement a method for modeling preferences of customers to drive personalized contextual recommendations using cognitive deep constrained filtering comprising:
• receiving, by processor, a user query input into a front end interface generated by an online retail platform, the front end user interface transmitted to a user computing device by the online retail platform to enable interaction between a user and the online retail platform, wherein the user is identified upon accessing the online retail platform;
• in response to receiving the user query, performing, by the processor, a first online matrix manipulation and a second online matrix manipulation,
• wherein the first online matrix manipulation includes:
• querying a customer similarity matrix while the user is interacting with the front end interface to identify customers most similar to the user, the customer similarity matrix created by the processer communicating with a plurality of neighboring computing devices associated with neighbors of the user to determine a number of neighbors that are similar to the user;
• creating a user feature matrix and a product features matrix specific to the user and a product in the user query, respectively;
• storing the product features matrix on a database connected to the computing system over a network;
• and
• obtaining a sparse data matrix based on the customer similarity matrix, the user feature matrix, and the products feature matrix;
• wherein the second online matrix manipulation includes:
• creating a product similarity matrix based on a candidate set of attributes extracted from the user query,
• filtering the results of the sparse data matrix with the product similarity matrix to determine a list of ranked recommended products for the user;
• training, by the processor, a model with a deep neural network using data from the product similarity matrix and data from the product feature matrix as inputs to obtain a trained model;
• inputting, by the processor, a new product information into the trained model to map product similarity and item features of existing products to new products to predict product features for new products to be used with the product feature matrix;
• updating, by the processor, the database storing the product features matrix to revise the product features matrix over time as new products become available;
• and
• sending, by the processor, the list of ranked recommended products to the front end user interface of the user computing device.

A computer program product, comprising a computer readable hardware storage device storing a computer readable program code, the computer readable program code comprising an algorithm that when executed by a computer processor of a computing system implements a method for modeling preferences of customers to drive personalized contextual recommendations using cognitive deep constrained filtering, comprising:

• receiving, by the processor, a user query input into a front end interface generated by an online retail platform, the front end user interface transmitted to a user computing device by the online retail platform to enable interaction between a user and the online retail platform, wherein the user is identified upon accessing the online retail platform;
• in response to receiving the user query, performing, by the processor, a first online matrix manipulation and a second online matrix manipulation,
• wherein the first online matrix manipulation includes:
• querying a customer similarity matrix while the user is interacting with the front end interface to identify customers most similar to the user, the customer similarity matrix created by the processer communicating with a plurality of neighboring computing devices associated with neighbors of the user to determine a number of neighbors that are similar to the user;
• creating a user feature matrix and a product features matrix specific to the user and a product in the user query, respectively;
• storing the product features matrix on a database connected to the computing system over a network;
• and
• obtaining a sparse data matrix based on the customer similarity matrix, the user feature matrix, and the products feature matrix;
• wherein the second online matrix manipulation includes:
• creating a product similarity matrix based on a candidate set of attributes extracted from the user query,
• filtering the results of the sparse data matrix with the product similarity matrix to determine a list of ranked recommended products for the user;
• training, by the processor, a model with a deep neural network using data from the product similarity matrix and data from the product feature matrix as inputs to obtain a trained model;
• inputting, by the processor, a new product information into the trained model to map product similarity and item features of existing products to new products to predict product features for new products to be used with the product feature matrix;
• updating, by the processor, the database storing the product features matrix to revise the product features matrix over time as new products become available;
• and
• sending, by the processor, the list of ranked recommended products to the front end user interface of the user computing device.