Recommendation techniques for target execution
Multi-modal techniques are described for identifying a recommendation corresponding to a system-defined target strategy. A machine-learning model configured to generate output data from input data may be trained or otherwise obtained. The machine-learning model being trained using an unsupervised machine-learning algorithm and a data set comprising a recurrent data instance, a historical user data instance, and a catalog data instance, to generate output data comprising one or more classification labels for the one or more data instances. Using those classification labels and underlying data, a recommendation for implementing a system-defined target strategy can be identified and presented at a user interface. Upon receiving user approval, a number of corresponding operations may be executed to implement the system-defined target strategy.
1 . A computer-implemented method, comprising:
training, by a computing system, a machine-learning model to generate output data from input data, the machine-learning model being trained to assign a classification label that indicates a threshold degree of similarity between data instances, the machine-learning model being trained based at least in part on an unsupervised machine-learning algorithm and a training data set including data instances comprising a recurrent data instance that indicates a first set of items that were historically presented at a third-party website, a historical user data instance that indicated a second set of items that are associated with a historical target strategy, and a catalog data instance that indicates a metric corresponding to a third set of items that were previously provided at an electronic catalog, the output data comprising a respective classification label corresponding to a subset of the data instances of the training data set;
receiving, by the computing system, subsequent input data associated with a user account, the user account being associated with a seller of items, the input data comprising attributes of one or more user-defined target strategies;
identifying, based at least in part on the classification label being assigned by the machine-learning model to the subsequent input data and one or more data instances, a recommendation for implementing a system-defined target strategy, the recommendation being associated with adding an item to a catalog that is associated with the seller or featuring the item in subsequent content, the item being selected from one or more items indicated by the one or more data instances that correspond to the classification label;
presenting the recommendation at a user interface;
receiving, at the user interface, user input indicating selection of the recommendation;
presenting, via a second user interface, one or more attributes corresponding to the system-defined target strategy, the one or more attributes identifying the item;
receiving subsequent user input indicating a selection of an option corresponding to the second user interface; and
executing, by the computing system based at least in part on receiving the subsequent user input, one or more operations associated with implementing the system-defined target strategy.
2 . The computer-implemented method of claim 1 , wherein the classification label is one of a plurality of classification labels.
3 . The computer-implemented method of claim 1 , wherein the recurrent data instance corresponds to data that identifies recurrent topics occurring at one or more third-party websites.
4 . The computer-implemented method of claim 1 , wherein the historical user data instance includes corresponding attributes associated with a previously known target strategy associated with a second user.
5 . The computer-implemented method of claim 4 , wherein the one or more attributes of the system-defined target strategy are identified from the previously known target strategy based at least in part on identifying a respective metric corresponding to the previously known target strategy exceeds a predefined threshold.
6 . The computer-implemented method of claim 1 , wherein the catalog data instance comprises corresponding attributes associated with the third set of items.
7 . The computer-implemented method of claim 1 , wherein the one or more operations comprise at least one of: 1) presenting content featuring a first item identified from the output data generated by the machine-learning model or 2) adding a second item to the electronic catalog on behalf of an entity corresponding to the user account.
8 . A computing device, comprising
one or more processors; and
one or more memories comprising computer-readable instructions that, when executed by the one or more processors, causes the computing device to:
obtain a machine-learning model configured to generate output data from input data, the machine-learning model being trained to assign a classification label that indicates a threshold degree of similarity between data instances based at least in part on an unsupervised machine-learning algorithm and a training data set comprising a recurrent data instance that identifies a first item historically presented at a third-party website, a historical user data instance that identifies a second item that is associated with a historical target strategy, and a catalog data instance that identifies a metric corresponding to a third item that was previously provided in an electronic catalog, the output data comprising a respective classification label corresponding to a subset of the data instances of the data set;
receive subsequent input data associated with a user account, the user account being associated with a seller of items, the input data comprising attributes of one or more user-defined target strategies;
identify, based at least in part on the classification label being assigned by the machine-learning model to the subsequent input data and one or more data instances of the data set, a recommendation for implementing a system-defined target strategy, the recommendation being associated with adding an item to a catalog that is associated with the seller or featuring the item in subsequent content, the item being selected from one or more items indicated by the one or more data instances that correspond to the classification label;
present the recommendation at a user interface;
receive, at the user interface, user input indicating selection of the recommendation; and
execute, based at least in part on receiving subsequent user input, one or more operations associated with implementing the system-defined target strategy.
9 . The computing device of claim 8 , wherein executing the instructions further causes the computing device to:
present, via a second user interface, one or more system-defined attributes corresponding to the system-defined target strategy; and
receive additional user input indicating a selection of an option corresponding to the second user interface, wherein the one or more operations are executed further based at least in part on receiving the additional user input.
10 . The computing device of claim 8 , wherein the recurrent data instance is associated with a social media website and a topic occurring over a threshold number of times at the social media website.
11 . The computing device of claim 8 , wherein the historical user data instance corresponds to high-performing user account associated with a service provider.
12 . The computing device of claim 11 , wherein the service provider hosts the electronic catalog comprising the catalog data instance.
13 . The computing device of claim 8 , wherein the one or more classification labels are ranked based at least in part on generating a score corresponding to a set of data instances associated with each classification label of the one or more classification labels.
14 . A non-transitory computer-readable storage medium comprising computer-readable instructions that, when executed by one or more processors of a computing device, cause the computing device to:
train a machine-learning model to generate output data from input data, the machine-learning model being trained to assign a classification label that indicates a threshold degree of similarity between data instances based at least in part on an unsupervised machine-learning algorithm and a training data set, the training data set comprising a recurrent data instance that identifies a first item that was historically presented at a third-party website, a historical user data instance that identifies a second item that is associated with a historical target strategy, and a catalog data instance that identifies a metric corresponding to a third item that was previously provided at an electronic catalog, the output data comprising a respective classification label corresponding to a subset of data instances of the training data set;
obtain historical target strategy data corresponding to one or more previously implemented and user-defined target strategies associated with a user account corresponding to a seller of items;
identify, based at least in part on the classification label being assigned by the machine-learning model to one or more data instances of the data set and the historical target strategy data, a recommendation for implementing a system-defined target strategy, the recommendation being associated with adding an item to a catalog that is associated with the seller or featuring the item in subsequent content, the item being selected from one or more items indicated by the one or more data instances that correspond to the classification label;
present the recommendation at a user interface;
receive, at the user interface, user input indicating selection of the recommendation; and
execute, based at least in part on receiving subsequent user input, one or more operations associated with implementing the system-defined target strategy.
15 . The non-transitory computer-readable storage medium of claim 14 , wherein at least one classification label is associated with at least one recurrent data instance, at least one historical user data instance, and at least one target strategy data instance.
16 . The non-transitory computer-readable storage medium of claim 14 , wherein the recurrent data instance, the historical user data instance, and the catalog data instance are obtained from one or more data provider computers.
17 . The non-transitory computer-readable storage medium of claim 16 , wherein the one or more data provider computers are associated with different entities.
18 . The non-transitory computer-readable storage medium of claim 14 , wherein the unsupervised machine-learning algorithm comprises a clustering algorithm configured to identify relationships between data instances of the input data.
19 . The non-transitory computer-readable storage medium of claim 14 , wherein executing the instructions further causes the computing device to:
compute one or more similarity measures between data instances of different classification labels; and
evaluate a degree of accuracy of the machine-learning model based at least in part on the one or more similarity measures.