| CPC G06Q 30/0631 (2013.01) [G06F 16/212 (2019.01); G06N 5/04 (2013.01); G06N 20/00 (2019.01); G06Q 30/0201 (2013.01)] | 20 Claims |
|
1. A system comprising:
one or more processors; and
one or more non-transitory computer-readable media storing computing instructions that, when executed on the one or more processors, cause the one or more processors to perform operations comprising:
training a re-ranking algorithm by utilizing a training data set comprising a set of recommendation triplets, a label set, and a set of product types as inputs and weights assigned to each product feature of each product type as outputs;
generating one or more item relational graphs for one or more items based on historical user purchases;
transforming, using spectral filtering, the one or more item relational graphs into one or more frequency signals to remove noise from the one or more frequency signals;
training a feed-forward neural network machine learning model using a data set based on input for the data set comprising historical item data over a first period of time by:
encoding, using contextual feature encoding, the one or more items, wherein a respective encoding for each of the one or more items encoded comprises a respective vector;
constructing, using the feed-forward neural network machine learning model, one or more item pair label classifications for one or more item pairs of the one or more items as output of the feed-forward neural network machine learning model, wherein input of the feed-forward neural network machine learning model comprises the one or more frequency signals, wherein the feed-forward neural network machine learning model labels the one or more item pairs based on levels of similarity between items of the one or more item pairs, and wherein each item of the one or more items is associated with a respective vector code;
generating a respective similarity score for each of the one or more item pairs;
outputting a top k results for the one or more item pairs ranked by the respective similarity scores for the one or more item pairs, wherein the top k results are configured to be displayed in an order on a user interface of an electronic device of a user as a first display;
combining data from a personalization layer with the top k results to output a personalized ordered list of items associated with an item of the one or more items;
re-ranking, using the re-ranking algorithm, as trained, the top k results of the one or more item pairs based on the personalized ordered list of items, wherein the re-ranking algorithm further re-ranks the order of the top k results of the first display based on the personalization layer; and
displaying a second display presenting the personalized ordered list of items on the user interface of the electronic device of the user based on an output of the re-ranking algorithm.
|