Abstract: Segment valuation techniques usable in a digital medium environment are described. To do so, a segment valuation system first identifies the attributes that are significant in achievement of a desired metric (e.g., conversion) and then values segments based on those significant attributes. Attributes are selected from the trained model based on significance of those attributes towards achieving the desired metric. A valuation of a segment may then be calculated based on the valuations of these attributes. For example, inclusion of the selected attributes within a segment, and the valuations of those selected attributes, is then used by the segment valuation system to generate data describing a value of the segment towards achieving the metric.

Abstract: A digital experience enhancement system includes an ensemble deep learning model that includes an estimator ensemble and a neural network. The ensemble deep learning model is trained to generate a digital experience enhancement recommendation from an enhancement request. The ensemble deep learning model receives the enhancement request, which is input to the estimator ensemble. The estimator ensemble uses various different machine learning systems to generate estimator output values. The neural network uses the estimator output values from the estimator ensemble to generate a digital experience enhancement recommendation. The digital experience generation system then uses this digital experience enhancement recommendation to enhance the digital experience.

Abstract: Systems, methods, and non-transitory computer-readable media (systems) are disclosed for generating meaningful and insightful user segment reports based on a high dimensional data space. In particular, in one or more embodiments, the disclosed systems utilize a relaxed bi-clustering model to automatically identify user segments in a data space including datasets of features specific to individual users. In at least one embodiment, the disclosed systems identify and include users in automatically generated user segments even though those users are associated with some, but perhaps not all, of the features as other members in the automatically generated user segments.

Abstract: Matrix completion and recommendation provision with deep learning is described. A matrix manager system imputes unknown values of incomplete input matrices using deep learning. Unlike conventional techniques, the matrix manager system completes incomplete input matrices using deep learning regardless of whether an input matrix represents numerical, categorical, or a combination of numerical and categorical attributes. To enable a machine-learning model (e.g., an autoencoder) to complete a matrix, the matrix manager system initially encodes the matrix. This involves normalizing known values of numerical attributes and categorically encoding known values of categorical attributes. The matrix manager system performs categorical encoding by replacing information of a given categorical attribute (e.g., an attribute column) with replacement information for each possible value of the attribute (e.g., new columns for each possible value).

Abstract: Techniques disclosed herein relate generally to evaluating and selecting candidate datasets for use by software applications, such as selecting candidate datasets for training machine-learning models used in software applications. Various machine-learning and other data science techniques are used to identify unique entities in a candidate dataset that are likely to be part of target entities for a software application. A merit attribute is then determined for the candidate dataset based on the number of unique entities that are likely to be part of the target entities, and weights associated with these unique entities. The merit attribute is used to identify the most efficient or most cost-effective candidate dataset for the software application.

Abstract: User data overlap determination in a digital medium environment is described. Initially, a user selects segments of user data for which a determination of overlap is to be made. For example, the user selects a segment representing users that are working professionals and a segment representing users that are mothers, such that working-mother users may correspond to the overlap. Regardless of the particular segments selected, an indication of those segments is received. One of multiple different overlap determining techniques—which can include a combined MinHash and HyperLogLog (HLL) technique and an Inclusion-Exclusion technique—may be selected for computing the overlap based on a number of segments indicated and numbers of users represented by the segments. The selected overlap determining technique is then used to compute the user data overlap between the indicated segments. Digital content including values indicative of the determined overlap is generated for presentation to a user.

Abstract: Techniques disclosed herein relate generally to evaluating and selecting candidate datasets for use by software applications, such as selecting candidate datasets for training machine-learning models used in software applications. Various machine-learning and other data science techniques are used to identify unique entities in a candidate dataset that are likely to be part of target entities for a software application. A merit attribute is then determined for the candidate dataset based on the number of unique entities that are likely to be part of the target entities, and weights associated with these unique entities. The merit attribute is used to identify the most efficient or most cost-effective candidate dataset for the software application.

Abstract: Categorical data transformation and clustering techniques and systems are described for machine learning using natural language processing. These techniques and systems are configured to improve operation of a computing device to support efficient and accurate use of categorical data, which is not possible using conventional techniques. In an example, categorical data is received by a computing device that includes a categorical variable having a non-numerical data type for a number of classes. The categorical data is then converted into numerical data using natural language processing. Data is then generated by the computing device that includes a plurality of latent classes. This is performed by clustering the numerical data into a number of clusters that is smaller than the number of classes in the categorical data.

Abstract: Techniques disclosed herein relate generally to evaluating and selecting candidate datasets for use by software applications, such as selecting candidate datasets for training machine-learning models used in software applications. Various machine-learning and other data science techniques are used to identify unique entities in a candidate dataset that are likely to be part of target entities for a software application. A merit attribute is then determined for the candidate dataset based on the number of unique entities that are likely to be part of the target entities, and weights associated with these unique entities. The merit attribute is used to identify the most efficient or most cost-effective candidate dataset for the software application.

Abstract: Embodiments of the present invention provide systems, methods, and computer storage media for facilitating generation of prediction models. In some embodiments, a predetermined number of parameter value sets is identified. Each parameter value set includes a plurality of parameter values that represent corresponding parameters within a time series model. The parameter values can be selected in accordance with stratified sampling to increase a likelihood of prediction accuracy. The parameter value sets are input into a time series model to generate a prediction value in accordance with observed time series data, and the parameter value set resulting in a least amount of prediction error can be selected and used to generate a time series prediction model (ARIMA, AR, MA, ARMA) with corresponding model parameters, such as p, q, and/or k, subsequently used to predict values.

Abstract: Segment valuation techniques usable in a digital medium environment are described. To do so, a segment valuation system first identifies the attributes that are significant in achievement of a desired metric (e.g., conversion) and then values segments based on those significant attributes. Attributes are selected from the trained model based on significance of those attributes towards achieving the desired metric. A valuation of a segment may then be calculated based on the valuations of these attributes. For example, inclusion of the selected attributes within a segment, and the valuations of those selected attributes, is then used by the segment valuation system to generate data describing a value of the segment towards achieving the metric.

Abstract: Segment valuation techniques usable in a digital medium environment are described. To do so, a segment valuation system first identifies the attributes that are significant in achievement of a desired metric (e.g., conversion) and then values segments based on those significant attributes. Attributes are selected from the trained model based on significance of those attributes towards achieving the desired metric. A valuation of a segment may then be calculated based on the valuations of these attributes. For example, inclusion of the selected attributes within a segment, and the valuations of those selected attributes, is then used by the segment valuation system to generate data describing a value of the segment towards achieving the metric.

Abstract: Systems, methods, and non-transitory computer-readable media (systems) are disclosed for generating meaningful and insightful user segment reports based on a high dimensional data space. In particular, in one or more embodiments, the disclosed systems utilize a relaxed bi-clustering model to automatically identify user segments in a data space including datasets of features specific to individual users. In at least one embodiment, the disclosed systems identify and include users in automatically generated user segments even though those users are associated with some, but perhaps not all, of the features as other members in the automatically generated user segments.

Abstract: Embodiments of the present invention provide systems, methods, and computer storage media for digital user identification across different devices, channels, and venues. Generally, digital interactions of a user can reveal a pattern of digital behavior that can be detected and assigned to the user, and a classifier can be learned to identify the user. Various types of digital interaction data may be utilized to identify a user, including device data, geolocation data associated with a user device, clickstream data or other attributes of web traffic, and the like. Anonymity can be provided by only utilizing behavioral-based user data. Digital interaction data can be encoded and fed into a multi-class classifier (e.g., deep neural network, support vector machine, random forest, k-nearest neighbors, etc.), with each user corresponding to a different class. New users can be detected and used to automatically grow a deep neural network to identify additional classes for the new users.

Abstract: Categorical data transformation and clustering techniques and systems are described for machine learning. These techniques and systems are configured to improve operation of a computing device to support efficient and accurate use of categorical data, which is not possible using conventional techniques. In an example, categorical data is received by a computing device that includes a categorical variable having a non-numerical data type for a number of classes. The categorical data is then converted into numerical data based on clustering used to generate a plurality of latent classes.

Abstract: Systems, methods, and non-transitory computer-readable media (systems) are disclosed for generating meaningful and insightful user segment reports based on a high dimensional data space. In particular, in one or more embodiments, the disclosed systems utilize a relaxed bi-clustering model to automatically identify user segments in a data space including datasets of features specific to individual users. In at least one embodiment, the disclosed systems identify and include users in automatically generated user segments even though those users are associated with some, but perhaps not all, of the features as other members in the automatically generated user segments.

Abstract: Techniques for predicting a number of links an email campaign recipient will open are described. Elements in a dataset related to an email campaign are modeled using a tree structure, where nodes of the tree represent features of each element. A mean squared error is computed of an outcome for each of the elements to determine a weight for each respective tree. The weights are then regularized by applying a penalty, such as an elastic net penalty, to each of the weights. Then, the weights are applied to each of the trees. A weighted average of all of the outcomes of the trees is calculated, where the weighted average represents a prediction of an outcome resulting from a set of feature values. The feature values correspond to the nodes of each of the trees.

Abstract: A subset of items that can be identified, promoted, or recommended to the user is determined based in part on rankings or other feedback that the user has given to other items in the set. The techniques discussed herein employ localized regularization to generate estimated values for the unknown values. Regularization refers to adding information into the system in order to generate the unknown values. This additional information of the system is an estimate, and is generated based on the known properties of the system. The techniques discussed herein employ localized regularization, which refers to estimating additional information based on the particular user for which the unknown values are being generated. In contrast to employing global regularization that treats all users in the system the same, the localized regularization discussed herein treats each user independently of the other users.

Abstract: Systems and methods for determining whether to associate new user information with an existing user are disclosed. One embodiment involves clustering users in a set of users into clusters based on similarities of personal or behavioral features of the users. The embodiment further involves receiving new user information relating to a user using a device that provides the new user information via a computer network. A best matching cluster of the clusters is identified based on similarity of personal or behavioral features of the new user information to personal or behavioral features of the best matching cluster. The embodiment compares the personal or behavioral features of the new user information with personal or behavioral features of an existing user in the best matching cluster to determine whether to associate the new user information with the existing user or to assign it as a new (previously non-existent and unknown) user.

Abstract: Performance-based digital content delivery in a digital medium environment is described. Initially, different items of a collection of digital content are delivered to a substantially equal number of users. The collection is then iteratively tested to identify which content item achieves a desired action (e.g., conversion) at a highest rate. During the iterative test, data describing user interaction with the delivered content is collected. Based on the collected data, measures of achievement are determined for the different content items. Measures of statistical guarantee are also computed that indicate an estimated accuracy of the achievement measures. Responsive to determining that a condition for ending the test has not yet occurred, an optimized allocation is computed for delivery of the content by applying one of multiple allocation optimization techniques. The particular technique applied is based on the condition for ending the test and a type of statistical guarantee associated with the test.