Recommendation systems and techniques are described that employ change point detection to generate recommendations for digital content. In one example, a change point detection technique is employed by a recommendation system to identify when a change point has occurred at a respective time step of a series of time steps. Detection of this change point may then be used by the recommendation system to reset the statistical model to address this change as well as generate a subsequent recommendation configured for exploration of reward distributions of the items of digital marketing content.

Techniques for optimizing machine-learned models based on dwell time of network-transmitted content items are provided. In one technique, impression data and selection data are used train a selection prediction model. For each impression, a dwell time associated with that impression is determined and compared to a skip time. If the dwell time is less than the skip time, then a first training label that indicates that the impression is skipped is associated with the impression. If the dwell time is greater than the skip time, then a second training label that indicates that the impression is not skipped is associated with the impression. These training labels are used to train a skip prediction model. The selection prediction model and the skip prediction model are used in a content item selection event to generate a score for each candidate content item. The scores are used to select a content item.

Off-policy evaluation of a new target policy is performed using historical data gathered based on a previous logging policy to estimate the performance of the target policy. An estimator may be used, wherein either a quality-based estimator or a quality-agnostic estimator is used to weight the difference between an observed reward in the historical data and an estimated reward generated by the target policy. A quality-agnostic estimator may be used to evaluate an importance weight according to a threshold. In such examples, when the importance weight exceeds the threshold, the quality-agnostic estimator clips the importance weight at the threshold, thereby providing an fixed upper bound irrespective of the quality of the reward predictor. In other examples, a quality-based estimator is used, in which an upper bound incorporates the quality of the reward predictor in order to modify an importance weight used by the estimator.

Systems and methods are disclosed that measure web activity bursts after ad broadcasts that may be sent to multiple persons. One system uses a cookie-less/cookie-optional, anonymous/personal-identification-not-required, method for web-based conversion tracking that will work on broadcast media systems such as television, and could also be applied to measuring spikes from email, radio, and other forms of advertising where an episodic ad event is broadcast to multiple parties, and where responses occur in a batch after the broadcast.

A system is disclosed for presenting advertisements for products and related products for a consumer based on the products being purchased.

Systems, methods, and apparatus for utilizing forward market pricing to facilitate operational decisions are disclosed. An example method may include interpreting an energy resource utilization requirement for a task system, and interpreting a behavioral data source, comprising entity behavioral data for an entity associated with the task system. The method may also include forecasting a forward market price for an energy credit resource in response to the energy resource utilization requirement and the behavioral data source. Then, in response to the forecast of the forward market price for the energy credit resource, adjusting an operation of the task system or executing a transaction.

An ad is placed in a movie, by analyzing inherent characteristics of the movie, analyzing viewed characteristics of the movie, analyzing viewer characteristics of a viewer of the movie, obtaining advertiser preferences for placement of the ad in the movie, determining costs of placing the ad in the movie based on the inherent characteristics of the movie, the viewed characteristics of the movie, the viewer characteristics and the advertiser preferences, and placing the ad in the movie in accordance with the inherent characteristics of the movie, the viewed characteristics of the movie, the viewer characteristics, the advertiser preferences and the determined costs.

At least some embodiments are directed to a system that receives data generated when point of sale terminals are used. The point of sale terminals are associated with multiple merchants. The system generates a quantitative data from the received data. The system computes a quality of service score of a first merchant based on the quantitative data. The system generates an item data structure storing values that correlate multiple users with multiple merchants at least based on the quality of service score. The system determines users that are shared by multiple merchants and computes a customer relevance value associated with the user and the first merchant. The system outputs a user-specific recommendation associated with a second merchant to a user computing device, wherein the user-specific recommendation is computed at least in part based on the customer relevance value associated with the first merchant.

Systems and methods directed to managing computer resource allocation by monitoring signals indicating demand for services utilizing computer resources are described. A method includes maintaining, for each first event of first events, historical registration data and respective parameter values of the first event and identifying, for a second event having an open registration status, respective parameter values of the second event, and registration data for the second event. The method includes computing a similarity score between the second event and each first event of the plurality of first events, based on the respective parameter values of the first event and the second event and the registration data of the second event and the historical registration data of the first event, generating, for the second event, a projected number of entities based on determined information and determining a ranking of the second event.

Disclosed are systems and methods for deploying a targeted advertisement. The systems and methods may include identifying a time period of expected decreased traffic. The targeted advertisement may be generated based on constraints set by a user. Targeted advertisement may be deployed during a predetermined time period prior to the time period of expected decreased traffic.