Abstract: Automatic machine-learning processes and systems for an online social network are described. During operation of the online social network, a system can automatically collect labeled training events, obtain snapshots of raw entity data associated with subjects from the collected training events, produce training examples by generating features for each training event using the snapshots of entity data and current entity data, and split the training examples into a training set and a test set. Next, the system can use a machine-learning technique to train a set of models and to select the best model based on one or more evaluation metrics using the training set. The system can then evaluate the performance of the best model on the test set. If the performance of the best model satisfies a performance criterion, the system can use the best model to predict responses for the online social network.

Abstract: Automatic machine-learning processes and systems for an online social network are described. During operation of the online social network, a system can automatically collect labeled training events, obtain snapshots of raw entity data associated with subjects from the collected training events, produce training examples by generating features for each training event using the snapshots of entity data and current entity data, and split the training examples into a training set and a test set. Next, the system can use a machine-learning technique to train a set of models and to select the best model based on one or more evaluation metrics using the training set. The system can then evaluate the performance of the best model on the test set. If the performance of the best model satisfies a performance criterion, the system can use the best model to predict responses for the online social network.

Abstract: During a targeting technique, features are extracted. Some features are associated with attributes in profiles of users of a social network (which facilitates interactions among the users), and others are associated with existing types of recommendations previously provided to the users in recommendations or otherwise associated with the users. Then, relevancy scores are determined based on the extracted features. Moreover, one or more of the extracted features are selected as rules for identifying a subset of types of recommendations to target at the users.