Optimize shopping route using purchase embeddings
Aspects described herein may relate to methods, systems, and apparatuses that provide new capabilities for recommending purchases to a user based on the user's past purchasing history and the purchase history of others. A new descriptor referred to as “purchase embeddings” is disclosed, which are data records in a new multi-dimensional space for describing and tracking purchases of goods and services.
1 . A method, comprising:
receiving purchase information relating to a first plurality of items, wherein the purchase information comprises data records in a multi-dimensional space;
generating, using a classifier model, purchase embeddings from the data records in the multi-dimensional space;
identifying, based on the purchase information, a correlation between purchases of the first plurality of items associated with a first geographic region;
detecting a second geographic region;
identifying a second plurality of items;
calculating one or more measures of similarity between the first and the second pluralities of items;
identifying, using a machine learning model and based on the purchase embeddings and the calculating, a subset of the second plurality of items that are available for purchase within the second geographic region with and the one or more measures of similarity to the first plurality of items;
generating, based on the purchase embeddings and using the machine learning model, a proposed pattern for purchasing the subset within the second geographic region, wherein the proposed pattern provides one or more vendor locations within the second geographic region at which the subset is sold;
sending instructions comprising the proposed pattern;
receiving purchase information relating to the proposed pattern; and
updating the machine learning model based on the received purchase information and the generated proposed pattern.
2 . The method of claim 1 , further comprising limiting the one or more vendor locations based on a predetermined distance from the second geographic region and a distance from one or more other vendor locations.
3 . The method of claim 1 , further comprising:
identifying a possible location where one item in the subset may be purchased; and
selecting, based on one or more criteria, the possible location as one of the one or more vendor locations, wherein the one or more criteria include one or more of:
the possible location being identified as a small business,
a toll road being excluded in a route to the possible location,
the possible location being within a predefined distance to a location of a user,
the possible location being within a predefined distance to another one of the one or more vendor locations, or
the possible location being identical to another one of the one or more vendor locations.
4 . The method of claim 1 , further comprising:
identifying possible locations where one item of the subset may be purchased;
receiving, via a device, a user preference; and
selecting, based on the user preference, one of the possible locations as one of the one or more vendor locations.
5 . The method of claim 1 , further comprising:
comparing the proposed pattern to the purchase information; and
evaluating, based on the comparing, the proposed pattern according to one or more metrics, wherein the sending of the instructions is based on the proposed pattern exceeding a predetermined rating according to the one or more metrics.
6 . The method of claim 5 , wherein the proposed pattern exceeding the predetermined rating indicates a travel time, a travel distance, a purchase cost, or a sales tax being lower for the subset than for the first plurality of items.
7 . The method of claim 1 , further comprising ranking the measures of similarity between the first and the second pluralities of items.
8 . The method of claim 1 , wherein the second plurality of items are identified based on one or more of:
one or more geographic regions where one of the second plurality of items may be purchased,
a vendor location where one of the second plurality of items may be purchased,
a vendor identified as a small business where one of the second plurality of items may be purchased,
a class of goods or services of one of the second plurality of items, a similar item to one of the second plurality of items, or
an alternate name for one of the second plurality of items.
9 . The method of claim 1 , further comprising:
identifying a first time frame when the purchases of the first plurality of items occurred; and
including, based on the first time frame, a second time frame in the proposed pattern of when to purchase the subset.
10 . The method of claim 1 , wherein the correlation between the purchases of the first plurality of items is based on one or more of:
a common class of goods or services of two of the first plurality of items,
distances traveled for the purchases of the first plurality of items,
amounts paid for the purchases of the first plurality of items,
geographic areas where the purchases of the first plurality of items occurred,
time frames when the purchases of the first plurality of items occurred,
frequencies of repeated purchases of the first plurality of items,
durations between two of the purchases of the first plurality of items,
distances between where the purchases of the first plurality of items occurred, or
methods of payment for the purchases of the first plurality of items.
11 . The method of claim 1 , wherein the multi-dimensional space comprises a tuple.
12 . An apparatus, comprising:
at least one computer processor; and
computer memory comprising computer-executable instructions that when executed by the at least one computer processor, cause the apparatus to:
receive purchase information relating to a first plurality of items, wherein the purchase information comprises data records in a multi-dimensional space;
generate, using a classifier model, purchase embeddings from the data records in the multi-dimensional space;
identify, based on the purchase information, a correlation between purchases of the first plurality of items associated with a first geographic region;
detect a second geographic region;
identify a second plurality of items;
calculate one or more measures of similarity between the first and the second pluralities of items;
identify, using a machine learning model and based on the purchase embeddings and the calculating, a subset of the second plurality of items that are available for purchase within the second geographic region with and the one or more measures of similarity to the first plurality of items;
generate, based on the purchase embeddings and using the machine learning model, a proposed pattern for purchasing the subset within the second geographic region, wherein the proposed pattern provides one or more vendor locations within the second geographic region at which the subset is sold;
send instructions comprising the proposed pattern;
receive purchase information relating to the proposed pattern; and
update the machine learning model based on the received purchase information and the generated proposed pattern.
13 . The apparatus of claim 12 ,
wherein the proposed pattern provides one or more vendor locations within the second geographic region at which the subset is sold; and
wherein the computer-executable instructions, when executed by the at least one computer processor, further cause the apparatus to:
limit the one or more vendor locations based on a predetermined distance from the second geographic region and a distance from one or more other vendor locations.
14 . The apparatus of claim 12 , wherein the computer-executable instructions, when executed by the at least one computer processor, further cause the apparatus to:
compare the proposed pattern to the purchase information; and
evaluate, based on the comparing, the proposed pattern according to one or more metrics, wherein the instructions are based on the proposed pattern exceeding a predetermined rating according to the one or more metrics.
15 . The apparatus of claim 14 , wherein the proposed pattern exceeding the predetermined rating indicates a travel time, a travel distance, a purchase cost, or a sales tax being lower for the subset than for the first plurality of items.
16 . The apparatus of claim 12 , wherein the computer-executable instructions, when executed by the at least one computer processor, further cause the apparatus to rank the measures of similarity between the first and the second pluralities of items.
17 . The apparatus of claim 12 , wherein the second plurality of items are identified based on one or more of:
one or more geographic regions where one of the second plurality of items may be purchased,
a vendor location where one of the second plurality of items may be purchased,
a vendor identified as a small business where one of the second plurality of items may be purchased,
a class of goods or services of one of the second plurality of items, a similar item to one of the second plurality of items, and
an alternate name for one of the second plurality of items.
18 . The apparatus of claim 12 , wherein the computer-executable instructions, when executed by the at least one computer processor, further cause the apparatus to:
identify a first time frame when the purchases of the first plurality of items occurred; and
include, based on the first time frame, a second time frame in the proposed pattern of when to purchase the subset.
19 . A method, comprising:
receiving purchase information, from a device, of a first plurality of items;
generating, using a classifier model, purchase embeddings corresponding to multi-dimensional data records for the first plurality of items;
identifying a correlation between the purchases of the first plurality of items;
identifying a second geographic region that includes a location of the device;
determining a second plurality of items;
calculating one or more measures of similarity between the first and the second pluralities of items;
identifying, using a machine learning model and based on the purchase embeddings, a subset of the second plurality of items within the second geographic region;
generating, based on the subset and the purchase embeddings, and using the machine learning model, a proposed pattern for purchasing the subset within the second geographic region;
sending, to the device, instructions with the proposed pattern for purchasing the subset;
receiving purchase information relating to the proposed pattern for purchasing the subset; and
updating the machine learning model based on the received purchase information and the generated proposed pattern.
20 . The method of claim 19 , further comprising:
comparing the proposed pattern to the purchase information; and
evaluating, based on the comparing, the proposed pattern according to one or more metrics, wherein the sending of the instructions to the device is based on the proposed pattern exceeding a predetermined rating according to the one or more metrics.