Facebook, Inc.

Airbnb, Inc.

Haibin Cheng

Xiao Meng Zhang

Gang Wu

Boris Pierre Arnoux

Anh Phuong Bui

Malay Haldar

Dominique Marcel Georgette Van De Sompel

Sue Ann Hong

Xingyao Ye

Xian Xu

Tao Xu

Yang Pei

Brendan Marshall Collins

An online reservation system is configured to receive requests from a guest for searching property listings and to return property listings that satisfy the search criteria of the requests. The online reservation system uses a machine learning system to rank the property listings returned by the search. The machine learning system uses objective functions to determine parameters for each property listing and assign a ranking based on the parameters. A first objective function generates a parameter indicating an extent to which a property listing matches preferences of the guest, and is based on data about the guest's interactions with the reservation system. A second objective function generates another parameter indicating an extent to which the search request matches the preferences of the host associate with the property listing, and is based on data about the host's responses to reservation requests.

Ranking property listing search results

A social networking system presents users with a content items and ad requests, which may include targeting criteria specifying a topic. Interactions by users who were presented with an advertisement from an ad request including targeting criteria specifying the topic are stored by the social networking system and used to identify a cluster group of additional users having characteristics similar to characteristics of users who were presented with the advertisement from the ad request including targeting criteria specifying the topic and who interacted with the advertisement. The social networking system determines scores for additional users in the cluster group based on measures of similarity between the additional users and the users who were presented with the advertisement and who interacted with the advertisement. Based on the determined scores, the social networking system associates additional users in the cluster group with the topic.

Clustering users of a social networking system based on user interactions with content items associated with a topic

An advertising system identifies users associated with an interest topic and generates a list of such users in which all advertising accounts are proportionately represented in the list. Such users are identified by recording user-page access data to each page in a cluster of pages associated with the interest topic. A list of user-account associations is generated by grouping the user-page access data by the advertising account associated with each page. The list is then optimized so a proportion of user-account associations for each advertising account is less than or equal to a predetermined threshold. This ensures that no one advertising account is overrepresented in the list. Using the optimized list, the advertising system can target users associated with the list with advertisements related to the interest topic.

Advertisement targeting for an interest topic

An online system selects advertisements for a user based on characteristics of the user. The online system presents advertisements to the user having targeting criteria satisfied by the characteristics of the user. To increase the number of users eligible to be presented with an advertisement, the online system increases the users eligible to be presented with the advertisement to include users that do not meet targeting criteria included in the advertisement. The online system obtains a percentile of users based on a performance metric associated with the advertisement and determines a cutoff measure of affinity based on the percentile and measures of affinity between various users and the advertisement. A user is eligible to be presented with the advertisement if a measure of affinity between the user and the advertisement is greater than the cutoff measure of affinity for the advertisement.

Expansion of targeting criteria using an advertisement performance metric to maintain revenue

Lookalike models can select users that are predicted to share characteristics with a specified set of seed users. The processing requirements for lookalike models can be decreased by identifying features that have low impact on model accuracy, and therefore can be excluded from creating models. Also, by identifying preferred seed sources and training parameters, accurate lookalike models can be created with less overhead and in less time. The features and training parameters can be identified by obtaining a sample seed set, extracting seeds with a defined set of features, and using the remaining training seeds to train a model. Performance of this model can be compared to a standard model to see if the model performs well. If so, features excluded from the features used to create the model, a seed source, or training parameters used to create the model can be selected.

Lookalike evaluation

An online system selects advertisements for presentation a user based on characteristics of the user. The online system monitors performance of advertisements based on a goal for the advertisement and a time interval for achieving the goal. During a time period within the time interval, the online system determines an actual performance of the advertisement and compares the actual performance to a portion of the goal associated with the time period. If the actual performance does not satisfy the portion of the goal associated with the time period, the online system expands targeting criteria of the advertisement to increase a number of users eligible to be presented with the advertisement.

Growing community of inventors

San Jose, CA, United States of America

Haibin Cheng