| CPC G06Q 10/087 (2013.01) | 20 Claims |

|
1. A computer-implemented method comprising:
compiling, using a database, an inventory list for a plurality of websites comprising a first website and a second website;
detecting, using a browser extension on a user device associated with a user, user interaction information comprising a cursor selecting a first product out of stock at the first website;
collecting, using the browser extension on the user device, product information associated with one or more attributes of the first product displayed in the first website;
determining, based on the product information, that the first product is in the inventory list;
providing, as input to a first machine learning model, the user interaction information;
determining, based on output from the first machine learning model, a likelihood that the user would navigate away from the first website;
after determining that the likelihood does not exceed a threshold value:
monitoring further user interaction information associated the first product displayed in the first website;
providing, as input to the first machine learning model, the further user interaction information; and
updating, based on output from the first machine learning model, the likelihood that the user would navigate away from the first website; and
after determining that the likelihood exceeds the threshold value and before the user navigates away from the first website:
processing the product information to generate a record comprising a plurality of key words;
converting, based on the key words, the record into text embeddings or image embeddings corresponding to a vector of features;
determining, using a second machine learning model and based on the vector of features, a second product in the inventory list that is similar to the first product displayed in the first website, wherein the second product is in stock at the second website; and
causing, using the browser extension on the user device, display of the second product in an overlay window displayed in the first website.
|