Microsoft Technology Licensing, LLC

Ashis Kumar Roy

David Maxwell Chickering

Lawrence Andrew Koch

David Earl Heckerman

Christopher A Meek

Alan Geller

Phani Vaddadi

David Max Chickering

Robert Paul Gorman

Prasanth Pulavarthi

Jeffery C Beman

Manan Sanghi

Izzet Can Envarli

Ole Christian Hellevik

Kiran Kumar Govindaraju Bangalore

An automatic rule generation and evaluation system enables a user to generate rules and evaluate the rules even as the rules are generated. The user employs a rules workbench user interface (UI) to provide input regarding the rules to be implemented on data obtained from a plurality of data sources. The rules workbench UI generates scripts from the user input. The scripts are employed to build a model and generate queries. Displaying the result sets obtained from executing the queries enables the user to evaluate the rules in real-time. If the user is satisfied, the user may finalize the rules, else the user can further modify the rules and the steps of rule modification and rule evaluation may be iterated until the rules are finalized.

Automated code generation for computer software

A system and method for estimating available payload inventory are provided. An advertisement delivery system generates a set of atomic market segment arrays from target market criteria for one or more advertisement campaigns to be served. The set of arrays is incremented corresponding to advertisement requests matching the target market criteria. The atomic market segment is processed to select an advertisement and to predict future capacity and manage inventory.

System and method for estimating available payload inventory

Computer-readable media for determining whether to accept a candidate order from a content provider, or advertiser, to display a particular number of advertisements within a specified time segment are provided. Initially, the content provider may include placement criteria that, among other things, identify a leaf node at which impressions of the advertisement are expected to be rendered. Generally, the leaf node refers to a location within a topic graph that describes inventory that is permissible to allocate to satisfy the candidate order. To perform the determination, the inventory of impressions available for accommodating the candidate order and a log of booked orders scheduled to be placed within the time segment are identified. Linear programs are then utilized to determine whether the estimated inventory that satisfies the placement criteria is available by predictively placing the booked orders at the estimated inventory. If estimated inventory remains available, the candidate order is accepted.

Linear-program formulation for optimizing inventory allocation

Embodiments relate to generating and using a directed graph as an advertising network taxonomy. Individual delivery locations provided by publishers for presenting advertisements are identified. A directed graph is generated that includes specific advertising opportunities that each correspond with an individual delivery location. The directed graph also includes multiple general advertising opportunities that correspond with aggregations of delivery locations. Multiple aggregations may each include all or part of the same set of delivery locations. Advertisers may then use the directed graph to purchase advertising opportunities, and advertisements may be served to delivery locations based on the advertiser selections.

Using a directed graph as an advertising system taxonomy

Enhancements to Bayesian prediction models for network location traffic provide increased accuracy in web traffic predictions. The enhancements include implementing user advertising target queries to determine preferred edges of a Bayesian model, employing hierarchical data structures to cleanse training data for a Bayesian model, and/or augmenting existing data with new training data to enhance a previously constructed Bayesian model. Preferred edge enhancements for the Bayesian model utilize target attribute derived preferred edges and/or explicitly specified preferred edges. Training data is cleansed utilizing tag hierarchies that can employ parent-child relationships, ancestor relationships, and/or network location specific parameters. New training data can also be employed to adjust probabilities in a previously constructed Bayesian model. The new training data can be weighted differently than data represented by the previously constructed Bayesian model.

Bayesian probability accuracy improvements for web traffic predictions

Data slices of historical time series are leveraged to facilitate in more accurately predicting like data slices of future time series. Different predictive models are employed to detect outliers in different data slices to enhance the accuracy of the predictions. The data slices can be temporal and/or non-temporal attributes of a data set represented by the historical time series. In this manner, for example, a historical time series for a network location can be sliced temporally into one hour time periods as a function of a day, a week, a month, a year, etc. Outliers detected in these data slices can then be mitigated utilizing the predictive time series model by replacing the outlier with the expected value. The mitigated historical time series can then be employed in a predictive model to predict future web traffic for the network location (and advertising revenue values) with a substantial increase in accuracy.

Web data outlier detection and mitigation

Random samples without replacement are extracted from a distributed set of items by leveraging techniques for aggregating sampled subsets of the distributed set. This provides a uniform random sample without replacement representative of the distributed set, allowing statistical information to be gleaned from extremely large sets of distributed information. Subset random samples without replacement are extracted from independent subsets of the distributed set of items. The subset random samples are then aggregated to provide a uniform random sample without replacement of a fixed size that is representative of a distributed set of items of unknown size. In one instance, a multivariate hyper-geometric distribution is sampled by breaking up the multivariate hyper-geometric distribution into a set of univariate hyper-geometric distributions. Individual items of a uniform random sample without replacement are then determined utilizing a normal approximation of the univariate hyper-geometric distributions and a finite population correction factor.

Growing community of inventors

Redmond, WA, United States of America

Ashis Kumar Roy