Dynamically determining customer intent and related recommendations using deep learning techniques
Methods, apparatus, and processor-readable storage media for dynamically determining customer intent and related recommendations using deep learning techniques are provided herein. An example computer-implemented method includes generating a prediction as to whether a user will order a particular offering during a user browsing session by applying one or more deep learning techniques to browsing session data derived from the user browsing session; mapping the user browsing session to one or more pre-established procurement user types by comparing the browser session data to data pertaining to multiple pre-established procurement user types; determining a recommendation of one or more offerings distinct from the particular offering, wherein determining the recommendation is based at least in part on the generated prediction and the mapping of the user browsing session to one or more of the multiple pre-established procurement user types; and outputting, within the user browsing session, the recommendation to the user.
1. A computer-implemented method comprising:
generating a prediction as to whether a user will order at least one particular offering during a user browsing session of one or more web pages of an enterprise, wherein generating the prediction comprises applying one or more deep learning techniques to browsing session data derived from the user browsing session, and wherein applying the one or more deep learning techniques to the browsing session data comprises processing data pertaining to a sequence of multiple browsing session events, occurring at distinct time instances, using multiple recurrent neural network cells and multiple hidden layers within a recurrent neural network, wherein one or more of the multiple hidden layers comprises a combination of input data at a given time instance and output data from at least one hidden layer associated with at least one previous time instance, and wherein processing comprises:
processing, using a first of the multiple recurrent neural network cells, data and corresponding temporal information pertaining to accessing at least one of the one or more web pages of the enterprise at a first time instance and a percentage of the at least one web page accessed by the user;
processing, using a second of the multiple recurrent neural network cells, data and corresponding temporal information pertaining to viewing the at least one particular offering on the at least one web page at a second time instance subsequent to the first time instance; and
processing, using a third of the multiple recurrent neural network cells, data and corresponding temporal information pertaining adding the at least one particular offering to a virtual shopping cart associated with the one or more web pages of the enterprise at a third time instance subsequent to the second time instance;
mapping the user browsing session to one or more of multiple pre-established procurement user types by comparing the browser session data to data pertaining to the multiple pre-established procurement user types, wherein comparing comprises calculating a distance from each of one or more data points within the browser session data to the mean data point associated with each the of multiple pre-established procurement user types using at least one multivariate support vector machine;
determining a recommendation of one or more offerings distinct from the at least one particular offering, wherein determining the recommendation is based at least in part on the generated prediction and the mapping of the user browsing session to one or more of the multiple pre-established procurement user types, and wherein determining the recommendation based at least in part on the mapping comprises identifying at least one offering purchased by at least one user associated with at least one of the multiple pre-established procurement user types upon a determination that at least one of the calculated distances, related to the at least one pre-established procurement user type, is less than a given amount; and
outputting, within the user browsing session, the recommendation to at least the user;
wherein the method is performed by at least one processing device comprising a processor coupled to a memory.
2. The computer-implemented method of claim 1 , wherein the browsing session data comprise data pertaining to time spent per offering.
3. The computer-implemented method of claim 1 , wherein the browsing session data comprise data pertaining to time spent per web page.
4. The computer-implemented method of claim 1 , wherein the browsing session data comprise data pertaining to one or more browsing frequency patterns.
5. The computer-implemented method of claim 1 , wherein the browsing session data comprise data pertaining to one or more return visits to one or more of the web pages.
6. The computer-implemented method of claim 1 , wherein the browsing session data comprise data pertaining to a manner in which the user accessed the one or more web pages of the enterprise.
7. The computer-implemented method of claim 1 , wherein the browsing session data comprise data pertaining to one or more offering reviews accessed during the user browsing session.
8. The computer-implemented method of claim 1 , wherein the browsing session data comprise data pertaining to one or more of the offerings selected for purchase during the user browsing session.
9. The computer-implemented method of claim 1 , wherein the one or more deep learning techniques are based at least in part on browsing session data derived from one or more previous user browsing sessions during which one or more users ordered at least one offering.
10. The computer-implemented method of claim 1 , wherein the one or more offerings distinct from the at least one particular offering comprises one or more offerings of a distinct offering category as compared to the at least one particular offering.
11. The computer-implemented method of claim 1 , wherein the one or more offerings distinct from the at least one particular offering comprises one or more offerings that are supplementary to the at least one particular offering.
12. The computer-implemented method of claim 1 , wherein determining the recommendation comprises identifying one or more offerings previously ordered by one or more users belonging to the one or more pre-established procurement user types mapped to the user browsing session.
13. A non-transitory processor-readable storage medium having stored therein program code of one or more software programs, wherein the program code when executed by at least one processing device causes said at least one processing device:
to generate a prediction as to whether a user will order at least one particular offering during a user browsing session of one or more web pages of an enterprise, wherein generating the prediction comprises applying one or more deep learning techniques to browsing session data derived from the user browsing session, and wherein applying the one or more deep learning techniques to the browsing session data comprises processing data pertaining to a sequence of multiple browsing session events, occurring at distinct time instances, using multiple recurrent neural network cells and multiple hidden layers within a recurrent neural network, wherein one or more of the multiple hidden layers comprises a combination of input data at a given time instance and output data from at least one hidden layer associated with at least one previous time instance, and wherein processing comprises:
processing, using a first of the multiple recurrent neural network cells, data and corresponding temporal information pertaining to accessing at least one of the one or more web pages of the enterprise at a first time instance and a percentage of the at least one web page accessed by the user;
processing, using a second of the multiple recurrent neural network cells, data and corresponding temporal information pertaining to viewing the at least one particular offering on the at least one web page at a second time instance subsequent to the first time instance; and
processing, using a third of the multiple recurrent neural network cells, data and corresponding temporal information pertaining adding the at least one particular offering to a virtual shopping cart associated with the one or more web pages of the enterprise at a third time instance subsequent to the second time instance;
to map the user browsing session to one or more of multiple pre-established procurement user types by comparing the browser session data to data pertaining to the multiple pre-established procurement user types, wherein comparing comprises calculating a distance from each of one or more data points within the browser session data to the mean data point associated with each the of multiple pre-established procurement user types using at least one multivariate support vector machine;
to determine a recommendation of one or more offerings distinct from the at least one particular offering, wherein determining the recommendation is based at least in part on the generated prediction and the mapping of the user browsing session to one or more of the multiple pre-established procurement user types, and wherein determining the recommendation based at least in part on the mapping comprises identifying at least one offering purchased by at least one user associated with at least one of the multiple pre-established procurement user types upon a determination that at least one of the calculated distances, related to the at least one pre-established procurement user type, is less than a given amount; and
to output, within the user browsing session, the recommendation to at least the user.
14. The non-transitory processor-readable storage medium of claim 13 , wherein determining the recommendation comprises identifying one or more offerings previously ordered by one or more users belonging to the one or more pre-established procurement user types mapped to the user browsing session.
15. The non-transitory processor-readable storage medium of claim 13 , wherein the one or more deep learning techniques are based at least in part on browsing session data derived from one or more previous user browsing sessions during which one or more users ordered at least one offering.
16. The non-transitory processor-readable storage medium of claim 13 , wherein the one or more offerings distinct from the at least one particular offering comprises at least one of one or more offerings of a distinct offering category as compared to the at least one particular offering and one or more offerings that are supplementary to the at least one particular offering.
17. An apparatus comprising:
at least one processing device comprising a processor coupled to a memory;
the at least one processing device being configured:
to generate a prediction as to whether a user will order at least one particular offering during a user browsing session of one or more web pages of an enterprise, wherein generating the prediction comprises applying one or more deep learning techniques to browsing session data derived from the user browsing session, and wherein applying the one or more deep learning techniques to the browsing session data comprises processing data pertaining to a sequence of multiple browsing session events, occurring at distinct time instances, using multiple recurrent neural network cells and multiple hidden layers within a recurrent neural network, wherein one or more of the multiple hidden layers comprises a combination of input data at a given time instance and output data from at least one hidden layer associated with at least one previous time instance, and wherein processing comprises:
processing, using a first of the multiple recurrent neural network cells, data and corresponding temporal information pertaining to accessing at least one of the one or more web pages of the enterprise at a first time instance and a percentage of the at least one web page accessed by the user;
processing, using a second of the multiple recurrent neural network cells, data and corresponding temporal information pertaining to viewing the at least one particular offering on the at least one web page at a second time instance subsequent to the first time instance; and
processing, using a third of the multiple recurrent neural network cells, data and corresponding temporal information pertaining adding the at least one particular offering to a virtual shopping cart associated with the one or more web pages of the enterprise at a third time instance subsequent to the second time instance;
to map the user browsing session to one or more of multiple pre-established procurement user types by comparing the browser session data to data pertaining to the multiple pre-established procurement user types, wherein comparing comprises calculating a distance from each of one or more data points within the browser session data to the mean data point associated with each the of multiple pre-established procurement user types using at least one multivariate support vector machine;
to determine a recommendation of one or more offerings distinct from the at least one particular offering, wherein determining the recommendation is based at least in part on the generated prediction and the mapping of the user browsing session to one or more of the multiple pre-established procurement user types, and wherein determining the recommendation based at least in part on the mapping comprises identifying at least one offering purchased by at least one user associated with at least one of the multiple pre-established procurement user types upon a determination that at least one of the calculated distances, related to the at least one pre-established procurement user type, is less than a given amount; and
to output, within the user browsing session, the recommendation to at least the user.
18. The apparatus of claim 17 , wherein determining the recommendation comprises identifying one or more offerings previously ordered by one or more users belonging to the one or more pre-established procurement user types mapped to the user browsing session.
19. The apparatus of claim 17 , wherein the one or more deep learning techniques are based at least in part on browsing session data derived from one or more previous user browsing sessions during which one or more users ordered at least one offering.
20. The apparatus of claim 17 , wherein the one or more offerings distinct from the at least one particular offering comprises at least one of one or more offerings of a distinct offering category as compared to the at least one particular offering and one or more offerings that are supplementary to the at least one particular offering.