International Business Machines Corporation

Other

Richard D Lawrence

Prem Melville

Yan Liu

Karthik Subbian

Wojciech Gryc

Estepan Meliksetian

Claudia Perlich

Dan Zhang

Ildar Khabibrakhmanov

Vijil Enara Chenthamarakshan

Upendra D Chitnis

Jingrui He

Cezar Pendus

Enara C Vijil

Edison Lao Ting

David C Gondek

Edwin Peter Dawson Pednault

Naoki Abe

Heng Cao

Vikas Sindhwani

Holly Edith Rushmeier

George S Almasi

David L Jensen

Grace Y Lin

James J Bennett

Yingdong Lu

Yunhee Jang

Vincent Philip Thomas

Arindam Banerjee

Roger R Gung

Shiva P Kasiviswanathan

Timothy R Bowden

Chandan Karrem Reddy

Saha Ankan

A method, system and computer program product are disclosed for predicting influence in a social network. In one embodiment, the method comprises identifying a set of users of the social network, and identifying a subset of the users as influential users based on defined criteria. A multitude of measures are identified as predictors of which ones of the set of users are the influential users. These measures are aggregated, and a composite predictor model is formed based on this aggregation. This composite predictor model is used to predict which ones of the set of users will have a specified influence in the social network in the future. In one embodiment, the specified influence is based on messages sent from the users, and for example, may be based on the number of the messages sent from each user that are re-sent by other users.

Predicting influence in social networks

Transfer learning is the task of leveraging the information from labeled examples in some domains to predict the labels for examples in another domain. It finds abundant practical applications, such as sentiment prediction, image classification and network intrusion detection. A graph-based transfer learning framework propagates label information from a source domain to a target domain via the example-feature-example tripartite graph, and puts more emphasis on the labeled examples from the target domain via the example-example bipartite graph. An iterative algorithm renders the framework scalable to large-scale applications. The framework propagates the label information to both features irrelevant to the source domain and unlabeled examples in the target domain via common features in a principled way.

Graph-based transfer learning

A system and method a Multi-Task Multi-View (MTV) learning problem. The method uses the label information from related tasks to make up for the lack of labeled data in a single task. The method further uses the consistency among different views to improve the performance. It is tailored for the above complicated dual heterogeneous problems where multiple related tasks have both shared and task-specific views (features), since it makes full use of the available information.

Graph-based framework for multi-task multi-view learning

A method, system and computer program product for inferring topic evolution and emergence in a set of documents. In one embodiment, the method comprises forming a group of matrices using text in the documents, and analyzing these matrices to identify a first group of topics as evolving topics and a second group of topics as emerging topics. The matrices includes a first matrix X identifying a multitude of words in each of the documents, a second matrix W identifying a multitude of topics in each of the documents, and a third matrix H identifying a multitude of words for each of the multitude of topics. These matrices are analyzed to identify the evolving and emerging topics. In an embodiment, the documents form a streaming dataset, and two forms of temporal regularizers are used to help identify the evolving topics and the emerging topics in the streaming dataset.

Inferring emerging and evolving topics in streaming text

A novel domain adaption/transfer learning method applied to the problem of classifying abbreviated documents, e.g., short text messages, instant messages, tweets. The method uses a large number of multi-labeled examples (source domain) to improve the learning on the partial observations (target domain). Specifically, a hidden, higher-level abstraction space is learned that is meaningful for the multi-labeled examples in the source domain. This is done by simultaneously minimizing the document reconstruction error and the error in a classification model learned in the hidden space using known labels from the source domain. The partial observations in the target space are then mapped to the same hidden space, and classified into the label space determined by the source domain.

System and method for domain adaption with partial observation

A computing system receives input data having both graph and unstructured data and computes a current log likelihood of the input data. The computing system compares the current log likelihood with a previous log likelihood of the input data. If the current log likelihood is larger than the previous log likelihood, the computing system update topic modeling parameters, community modeling parameters, and the link generation parameter until the computing system obtains a maximal value of the log likelihood of the input data. Then, the computing system creates a graph indicating topic similarity between the input data based on the topic modeling parameters, creates another graph indicating community similarity between entities associated with the input data based on the community modeling parameters, and predicts a link existence between input data or entities based on the link generation parameter, the topic modeling parameter and the community modeling parameter.

Systems and methods for extracting patterns from graph and unstructured data

Systems and methods for extracting patterns from graph and unstructered data

A method for automatically determining an Internet home page corresponding to a named entity identified by a specified descriptor including building a trained machine-learning model, generating candidate matches from the specified descriptor, wherein each candidate match includes an Internet address, extracting content-based features from websites associated with the Internet addresses of the candidate matches, determining a model score for each candidate match based on the content-based features using the trained machine-learning model, and determining a match from among the candidate matches according to the scores, wherein the match is returned as the Internet home page corresponding to the named entity.

Method and system using machine learning to automatically discover home pages on the internet

A method for identifying companies with specific business objectives that includes using existing sources of company firmographic data to identify a broad set of companies and associated websites, crawling the websites associated with the identified companies and indexing web site content for each of the identified companies with the specific business objective to realize indexed web content. The method further includes joining the company firmographic data with the indexed web content using a business objective common identifier to generate a store of joined structured firmographic data and indexed web content and presenting a display image representation of the store of joined structured firmographic data and indexed web content for user review. The display image further receives user input to score each of said companies identified therein, and using a search interface, querying the store of scored, joined structured firmographic data and indexed web content. The method further includes augmenting the search interface, or search results from a query, with predictive, machine-leaning processes that allow rapid identification of companies possibly missed in the query.

Method and system for identifying companies with specific business objectives

The present invention employs data processing systems to handle debt collection by formulation the collections process as a Markov Decision Process with constrained resources, thus making it possible automatically to generate an optimal collections policy with respect to maximizing long-term expected return throughout the course of a collections process, subject to constraints on the available resources possibly in multiple organizations. This is accomplished by coupling data modeling and resource optimization within the constrained Markov Decision Process formulation and generating optimized rules based on constrained reinforcement learning process comprising applied on the basis of past historical data.

Method and system for debt collection optimization

A method (and structure) for developing a distribution function for the probability of winning a bid by a seller for a product or service, using the seller's own historical data for winning bids and lost bids, includes normalizing the data for winning bids and the data for lost bids and merging the normalized data into a single set of data.

Method and structure for bid winning probability estimation and pricing model

A method and apparatus for retrieving and organizing data using a self-organizing map in a parallel transaction data base system. With this invention, input records of data are represented as n dimensional vectors. Each of these vectors then compressed by eliminating zeros in the components of these vectors. Then, the self-organizing map algorithm is applied to the compressed input records to group the records into a number of clusters, where each cluster comprises a number of records having common input parameters.

Scalable parallel algorithm for self-organizing maps with applications to sparse data mining problems

Growing community of inventors

Ridgefield, CT, United States of America

Richard D Lawrence