Provisioning interactive content based on predicted user-engagement levels

A system comprising one or more processors and a memory having stored thereon instructions that, upon execution by the one or more processors, cause the one or more processors to perform one or more operations, the system further comprising:

• a content generator configured to:
• transmit, to a user device, interactive content for a user, wherein the interactive content is associated with a targeted campaign initiated by a content provider system;
• receive, from a classifier subsystem, an output including a categorical value that represents a predicted user-engagement level of the user in response to a presentation of future interactive content associated with the targeted campaign;
• select a particular follow-up interactive content from a set of follow-up interactive content by using the categorical value as input, wherein resources allocated for generating the particular follow-up interactive content are determined in accordance with the categorical value;
• and
• transmit the particular follow-up interactive content to the user device, such that the user device displays the particular follow-up interactive content;
• and
• the classifier subsystem configured to:
• access user-activity data of the user, wherein the user-activity data includes a type of the interactive content and one or more user-device actions performed in response to the transmission of the interactive content to the user device;
• apply a machine-learning model to the user-activity data to generate the output that includes the categorical value that represents the predicted user-engagement level of the user at a particular future time point in response to a presentation of a future interactive content of the targeted campaign, wherein the particular future time point is defined by a preconfigured duration of time elapsed from a time when a particular user-device action is performed in response to the interactive content of the targeted campaign, wherein the machine-learning model was trained using a training dataset including previous user-device actions performed by a plurality of users in response to previous interactive content associated with other campaigns, and wherein the machine-learning model was trained at least by:
• identifying a time period within which the previous user-device actions were performed;
• splitting the time period into a set of time windows;
• training, for a subset of time windows of the set of time windows, the machine-learning model using a subset of the training dataset corresponding to a time window of the subset of time windows, wherein the subset of the training dataset includes previous user-device actions performed by at least one of the plurality of users and identified as being performed within the time window, wherein a duration of the time window is adjustable based on user-device actions in a corresponding subset of training dataset during training, wherein training the machine-learning model comprises creating a target label representing a known user-engagement level of a user associated with the subset of training dataset;
• and
• transmit the output to the content generator to trigger selection of the particular follow-up interactive content.

A computer program product tangibly embodied in a non-transitory machine-readable storage medium including instructions configured to cause one or more data processors to perform actions including:

• accessing user-activity data of a user, wherein the user-activity data includes a type of interactive content of a targeted campaign initiated by a content provider system and one or more user-device actions performed in response to presentation of the interactive content of the targeted campaign to the user;
• a step for applying a machine-learning model to the user-activity data to generate an output including a categorical value that represents a predicted user-engagement level of the user at a particular future time point in response to a presentation of future interactive content of the targeted campaign, wherein the particular future time point is defined by a preconfigured duration of time elapsed from a time when a particular user-device action is performed in response to the interactive content of the targeted campaign, wherein the machine-learning model was trained using a training dataset including previous user-device actions performed by a plurality of users in response to previous interactive content associated with other campaigns, and wherein the machine-learning model was trained at least by:
• identifying a time period within which the previous user-device actions were performed;
• splitting the time period into a set of time windows;
• and
• training, for a subset of time windows of the set of time windows, the machine-learning model using a subset of the training dataset corresponding to a time window of the subset of time windows, wherein the subset of the training dataset includes previous user-device actions performed by at least one of the plurality of users and identified as being performed within the time window, wherein a duration of the time window is adjustable based on user-device actions in a corresponding subset of training dataset during training, wherein training the machine-learning model comprises creating a target label representing a known user-engagement level of a user associated with the subset of training dataset;
• and
• selecting, based on the categorical value, a particular follow-up interactive content from a set of follow-up interactive content, wherein resources allocated for generating the particular follow-up interactive content are determined in accordance with the outputted categorical value.