Methods and systems for providing purchase recommendations based on responses to inquiries on product attributes
Systems and methods are disclosed for providing purchase recommendations. According to some examples, a method may include: determining respective frequency distributions of a plurality of vehicle attributes, the frequency distributions being determined based on occurrences of values of the plurality of vehicle attributes in a set of vehicles; selecting a vehicle attribute from the plurality of vehicle attributes based on the frequency distributions of the plurality of vehicle attributes; transmitting, to a user device, an inquiry for user preference regarding the selected vehicle attribute; receiving, from the user device, a response indicating the user preference; and presenting, to the user device, a recommendation of one or more vehicles determined based on the received response.
1. A computer-implemented method for determining recommendations based on user interaction, the computer-implemented method comprising:
determining a candidate set of vehicles;
iteratively adjusting the candidate set of vehicles by:
determining respective distributional characteristics of a plurality of vehicle attributes, each of the plurality of vehicle attributes occurring in the candidate set of vehicles as adjusted;
selecting a vehicle attribute from the plurality of vehicle attributes based on the respective distributional characteristics of the plurality of vehicle attributes;
transmitting, to a user device, an inquiry for user preference regarding the vehicle attribute;
receiving, from the user device, a response indicating the user preference;
adjusting the candidate set of vehicles based on the response;
determining whether to perform a next iteration; and
responsive to a determination not to perform the next iteration, determining a recommendation of one or more vehicles determined based on the candidate set of vehicles as adjusted.
2. The computer-implemented method of claim 1 , wherein selecting the vehicle attribute includes:
determining respective estimates of normalized entropy of the plurality of vehicle attributes based on the respective distributional characteristics; and
selecting the vehicle attribute from the plurality of vehicle attributes based on the respective estimates of normalized entropy.
3. The computer-implemented method of claim 2 , wherein
the respective distributional characteristics are percentage distributions, and
in determining the respective estimates of normalized entropy, the respective distributional characteristics are used as respective probability mass functions of the plurality of vehicle attributes.
4. The computer-implemented method of claim 2 , wherein selecting the vehicle attribute from the plurality of vehicle attributes includes selecting a vehicle attribute having a highest estimate of normalized entropy out of the respective estimates of normalized entropy of the plurality of vehicle attributes.
5. The computer-implemented method of claim 2 , wherein
the plurality of vehicle attributes includes a categorical attribute, and
for the categorical attribute, a respective estimate of normalized entropy is normalized based on a total number of values of the categorical attribute that appear at least once in the candidate set of vehicles.
6. The computer-implemented method of claim 1 , wherein
the plurality of vehicle attributes includes a quantitative attribute, and
determining the respective distributional characteristics includes determining a frequency distribution of the quantitative attribute by
grouping values of the quantitative attribute into bins, each of the values of the quantitative attribute corresponding to a respective vehicle in the candidate set of vehicles, and
determining a respective frequency of each of the bins, to obtain the frequency distribution of the quantitative attribute.
7. The computer-implemented method of claim 6 , wherein determining the frequency distribution of the quantitative attribute includes:
prior to grouping the values of the quantitative attribute into bins, processing the values of the quantitative attribute by capping one or more outlier values.
8. The computer-implemented method of claim 1 , wherein
the user preference indicates a preference for one or more values of the vehicle attribute,
adjusting the candidate set of vehicles based on the response includes excluding, from the candidate set of vehicles, vehicles having none of the one or more values for the vehicle attribute, to obtain a reduced set of vehicles as the candidate set of vehicles as adjusted, and
the recommendation includes the reduced set of vehicles as the one or more vehicles.
9. The computer-implemented method of claim 1 , wherein the plurality of vehicle attributes include one or more of: body style, exterior color, interior color, number of doors, vehicle condition, engine types, transmission type, fuel type, mileage, price, year, or fuel economy.
10. The computer-implemented method of claim 1 , wherein
in each subsequent iteration, vehicle attributes previously selected are omitted from further selections of vehicle attributes, and
subsequent distributional characteristics are recalculated for the candidate set of vehicles as adjusted.
11. The computer-implemented method of claim 1 , further comprising:
initiating a chat bot session enabling a user of the user device to communicate with a chat bot executed by a computer system,
wherein the inquiry is transmitted during the chat bot session, and the response is received during the chat bot session.
12. The computer-implemented method of claim 1 , further comprising:
presenting, to the user device, the recommendation of the one or more vehicles and one or more search filters of a vehicle listing interface, the one or more search filters having been determined in accordance with the user preference.
13. A computer system for determining recommendations based on user interaction, the computer system comprising:
a memory storing instructions; and
one or more processors configured to execute the instructions to perform a process, the process including:
determining respective distributional characteristics of a plurality of vehicle attributes, each of the plurality of vehicle attributes occurring in a set of vehicles;
selecting a vehicle attribute from the plurality of vehicle attributes based on the respective distributional characteristics of the plurality of vehicle attributes by
determining respective estimates of normalized entropy of the plurality of vehicle attributes based on the respective distributional characteristics using an equation η(X), where X is a variable having possible values x 1 , . . . , x n and corresponds to each of the plurality of vehicle attributes, the equation being:
η
(
X
)
=
-
∑
i
n
P
(
x
i
)
log
b
P
(
x
i
)
log
b
(
n
)
,
where P(x i ) is a probability for X=x i ,
b is an arbitrary logarithm base,
n is a count of the possible values x 1 , . . . , x n ; and
selecting the vehicle attribute from the plurality of vehicle attributes based on the respective estimates of normalized entropy;
transmitting, to a user device, an inquiry for user preference regarding the vehicle attribute;
receiving, from the user device, a response indicating the user preference; and
determining a recommendation of one or more vehicles determined based on the response.
14. The computer system of claim 13 , wherein, prior to determining the recommendation, the process is iterative,
in each subsequent iteration, vehicle attributes previously selected are omitted from further selections of vehicle attributes, and
subsequent distributional characteristics are recalculated for a candidate set of vehicles as adjusted based on user responses.
15. The computer system of claim 14 , wherein
the respective distributional characteristics are percentage distributions.
16. The computer system of claim 14 , wherein the process further includes, to select the vehicle attribute from the plurality of vehicle attributes, selecting a vehicle attribute having a highest estimate of normalized entropy out of the respective estimates of normalized entropy of the plurality of vehicle attributes.
17. The computer system of claim 14 , wherein
the plurality of vehicle attributes includes a categorical attribute, and
for the categorical attribute, the n is set to a total number of values of the categorical attribute that appear at least once in the set of vehicles.
18. The computer system of claim 14 , wherein
the plurality of vehicle attributes includes a quantitative attribute, and
the process further includes, to determine the respective estimates of normalized entropy, determining a frequency distribution of the quantitative attribute by
grouping values of the quantitative attribute into bins, each of the values of the quantitative attribute corresponding to a respective vehicle in the set of vehicles, and
determining a respective frequency of each of the bins, to obtain the frequency distribution of the quantitative attribute.
19. The computer system of claim 18 , wherein the process further includes, to determine the frequency distribution of the quantitative attribute:
prior to grouping the values of the quantitative attribute, processing the values of the quantitative attribute by capping one or more outlier values.
20. A computer system for implementing an artificial-intelligence interactive agent for determining recommendations based on user interaction, the computer system comprising:
a memory storing instructions; and
one or more processors configured to execute the instructions to perform a process, the process including:
determining respective distributional characteristics of a plurality of vehicle attributes describing vehicles in a set of vehicles, each of the plurality of vehicle attributes occurring in the set of vehicles;
determining respective estimates of normalized entropy of the plurality of vehicle attributes based on the respective distributional characteristics using an equation η(X), where X is a variable having possible values x 1 , . . . , x n and corresponds to each of the plurality of vehicle attributes, the equation being:
η
(
X
)
=
-
∑
i
n
P
(
x
i
)
log
b
P
(
x
i
)
log
b
(
n
)
,
where P(x i ) is a probability for X=x i ,
b is an arbitrary logarithm base,
n is a count of the possible values x 1 , . . . , x n ;
selecting a vehicle attribute from the plurality of vehicle attributes based on the respective estimates of normalized entropy;
transmitting, to a user device, an inquiry for user preference regarding the vehicle attribute;
receiving, from the user device, a response indicating the user preference; and
transmitting, to the user device, a web page including a list of one or more vehicles from the set of vehicles, the list of one or more vehicles having filters set in accordance with the user preference.