Automatic item grouping and personalized department layout system and method for reorder recommendations
A computer-implemented method including automatically generating predictions of a respective number of items that a user is likely to reorder in each of groups of the items that a user has ordered historically. The method also can include ranking the groups based on the predictions of the respective number of the items the user is likely to reorder in each of the groups. The method additionally can include transmitting for display to the user a user interface including the groups of the items. Other embodiments are described.
1 . A system comprising a processor and a non-transitory computer-readable medium storing computing instructions that, when executed on the processor, cause the processor to perform operations comprising:
automatically generating predictions of a respective number of unique items that a user computer, of a user, is likely to be used to reorder in each of groups of the unique items that the user has ordered historically, comprising:
traversing, by the processor, random forests of ensembles of decision trees for each unique item of the unique items, such that, at each decision point of the decision trees, a respective trained threshold parameter is used to determine a respective branch of each of the decision trees to follow until a respective leaf node of each of the decision trees is reached; and
determining a respective item prediction for each unique item of the unique items based on respective interim predictions obtained from the respective leaf nodes for each unique item of the unique items across the decision trees;
ranking the groups of the unique items based on the predictions of the respective number of the unique items the user is likely to reorder in each of the groups;
identifying a size of a display screen of the user computer;
adjusting, based on ranking the groups of the unique items and based on the size of the display, a webpage to include a scrolling module that is configured to display the groups of the unique items across multiple scrolling pages in a manner that decreases switching from the webpage to an item search interface; and
transmitting, for display on the display screen of the user computer, the adjusted webpage.
2 . The system of claim 1 , wherein the predictions automatically generated by the decision trees are averaged for each of the groups.
3 . The system of claim 1 , wherein the respective trained threshold parameter for each decision point of the decision trees is based on respective features comprising one or more of a number of orders, a recency of orders, a monetary value at a user level, a monetary value at a department level, reorder rates, similar item orders, or basket sizes.
4 . The system of claim 1 , wherein the groups are ordered in the user interface based on the ranking of the groups.
5 . The system of claim 1 , wherein the unique items are ordered within the groups based on a ranking of the unique items within the groups.
6 . The system of claim 5 , wherein the ranking of the unique items within the groups is based on reorder likelihood scores.
7 . The system of claim 1 , wherein the respective number of the unique items displayed within each of the groups is limited based on the respective number of the unique items that the user is likely to reorder in each of the groups.
8 . The system of claim 1 , wherein the interface webpage further comprises information identifying alternatives for out-of-stock items.
9 . A computer-implemented method comprising:
automatically generating predictions of a respective number of unique items that a user device, of a user, is likely to be used to reorder in each of groups of the unique items that the user has ordered historically, comprising:
traversing, by a processor, random forests of ensembles of decision trees for each unique item of the unique items, such that, at each decision point of the decision trees, a respective trained threshold parameter is used to determine a respective branch of each of the decision trees to follow until a respective leaf node of each of the decision trees is reached; and
determining a respective item prediction for each unique item of the unique items based on respective interim predictions obtained from the respective leaf nodes for each unique item of the unique items across the decision trees;
ranking the groups of the unique items based on the predictions of the respective number of the unique items the user is likely to reorder in each of the groups; and
identifying a size of a display screen of the user device;
adjusting, based on ranking the groups of the unique items and based on the size of the display, a user interface to include a scrolling module that is configured to display the groups of the unique items across multiple scrolling pages in a manner that decreases switching from the user interface to an item search interface; and
transmitting, for display on the display screen of the user device, information for the user interface.
10 . The computer-implemented method of claim 9 , wherein the predictions automatically generated by the decision trees are averaged for each of the groups.
11 . The computer-implemented method of claim 9 , wherein the respective trained threshold parameter for each decision point of the decision trees is based on respective features comprising one or more of a number of orders, a recency of orders, a monetary value at a user level, a monetary value at a department level, reorder rates, similar item orders, or basket sizes.
12 . The computer-implemented method of claim 11 , wherein the groups are ordered in the user interface based on the ranking of the groups.
13 . The computer-implemented method of claim 9 , wherein the unique items are ordered within the groups based on a ranking of the unique items within the groups.
14 . The computer-implemented method of claim 13 , wherein the ranking of the unique items within the groups is based on reorder likelihood scores.
15 . The computer-implemented method of claim 9 , wherein the respective number of the unique items displayed within each of the groups is limited based on the respective number of the unique items that the user is likely to reorder in each of the groups.
16 . The computer-implemented method of claim 9 , wherein the user interface further comprises information identifying alternatives for out-of-stock items.
17 . A non-transitory computer-readable medium storing computing instructions that, when executed on a processor, cause the processor to perform operations comprising:
automatically generating predictions of a respective number of unique items that a user device, of a user, is likely to be used to order in each of groups of the unique items that the user has ordered historically, comprising:
traversing, by the processor, random forests of ensembles of decision trees for each unique item of the unique items, such that, at each decision point of the decision trees, a trained threshold parameter is used to determine a respective branch of each of the decision trees to follow until a respective leaf node of each of the decision trees is reached; and
determining a respective item prediction for each unique item of the unique items based on respective interim predictions obtained from the respective leaf nodes for each unique item of the unique items across the decision trees;
ranking the groups of the unique items based on the predictions of the respective number of the unique items the user is likely to order in each of the groups;
identifying a size of a display screen associated with the user device;
adjusting, based on ranking the groups of the unique items and based on the size of the display, a page to include a scrolling module that is configured to display the groups of the unique items in a manner that decreases switching from the page to an item search interface; and
transmitting, for display on the display screen of the user device, information for the adjusted page.
18 . The non-transitory computer-readable medium of claim 17 , wherein the predictions automatically generated by the decision trees are averaged for each of the groups.
19 . The non-transitory computer-readable medium of claim 17 , wherein the respective trained threshold parameter for each decision point of the decision trees is based on respective features comprising one or more of a number of orders, a recency of orders, a monetary value at a user level, a monetary value at a department level, reorder rates, similar item orders, or basket sizes.
20 . The non-transitory computer-readable medium of claim 17 , wherein:
the groups are ordered in the page based on the ranking of the groups; and
the unique items are ordered within the groups based on a ranking of the unique items within the groups.