International Business Machines Corporation

Globalfoundries Inc.

Hongfei Li

Arun Hampapur

Dhaivat P Parikh

Wei Shan Dong

Zhiguo Li

Jinfeng Yi

Qing He

Wei Sun

Qi Yu

Anshul Sheopuri

Zhong Bo Jiang

Xuan Liu

Shilpa N Mahatma

Buyue Qian

Fei Wang

Hendrik F Hamann

Li Li

Jayant R Kalagnanam

Yada Zhu

Lexing Xie

Debarun Bhattacharjya

Andrew J Davenport

Frank D Fenhagen, Iv

Xiang Wang

Rick A Hamilton, Ii

Clifford A Pickover

Ning Duan

Chun Ma

Ajay Ashok Deshpande

Hongliang Fei

Linsong Chu

David Alexander Epstein

Perry G Hartswick

Songhua Xing

Ke Hu

Ching-Hua Chen-Ritzo

Ralph Peter Williams

Jing Xia

Karina Elayne Kervin

Srinivas Yarlanki

Dongping Fang

Michael Khayyat

A method, system and computer program product for identifying a geographic market share. Mobility data is acquired from applications running on mobile devices of users located within a geographic area. Mobility data is then used to infer shopping habits within the geographic area. Geo-demographic profiles are then created. The geographic market share is then determined using the created geo-demographic profiles.

Identifying geographic market share

A method is provided for personalized behavioral-driven event management. The method includes determining, by a processor, derivative behaviors from user behavioral data for one or more users. The derivative behaviors include intent driven behavioral data and habitual behavioral data. The method further includes deriving, by the processor, one or more behavioral triggers that differentiate the intent driven behavioral data from the habitual behavioral data. The method also includes generating, by the processor, a risk prediction based on a predictive risk assessment model that uses the intent driven behavioral data differentiated from the habitual behavioral data by the one or more behavioral triggers to enhance a predictive power of the user behavioral data. The method additionally includes enabling, by the processor, risk intervention for at least one of the one or more users by directly targeting the behavioral triggers, responsive to the risk prediction.

Personalized behavior-driven dynamic risk management with constrained service capacity

A method and system are provided. The method includes receiving by a computer having a processor and a memory, sequence data that includes labeled data and unlabeled data. The method further includes generating, by the computer having the processor and the memory, a recurrent neural network model of the sequence data, the recurrent neural network model having a recurrent layer and an aggregate layer. The recurrent neural network model feeds sequences generated from the recurrent layer into the aggregate layer for aggregation, stores temporal dependencies in the sequence data, and generates labels for at least some of the unlabeled data.

Customer profile learning based on semi-supervised recurrent neural network using partially labeled sequence data

An embodiment of the invention may include a method, computer program product and computer system for image identification and classification. The method, computer program product and computer system may include a computing device which may receive one or more images of a first object from at least two angles linguistic data associated with the first object. The computing device may input the one or more images of the first object into one or more first neural networks and the linguistic data of the first object into one or more second neural networks. The computing device may combine the output of the one or more first neural networks and the one or more second neural networks and generate an identification model based on the combined output of the one or more first neural networks and the one or more second neural networks.

Object recognition and description using multimodal recurrent neural network

Causal performance analysis for store merchandising may be provided. A clustering technique may be performed based on target store location data and existing store data. Based on the clustering technique, a peer selection group is determined comprising a group of stores determined to have similar attributes to the target store location. Sales distortions for a plurality of divisions associated with the group of stores in the peer selection group may be determined. A distortion matrix may be generated comprising a ranked list of the plurality of divisions. A merchandise mix recommendation for the target store location may be presented via a user interface device.

Causal performance analysis approach for store merchandizing analysis

A method and system are provided. The method includes converting, by a computer having a processor and a memory, categorical sequence data for a customer journey into a numerical similarity matrix. The method further includes learning, by the computer, features of the customer journey by applying a distance metric learning based matrix factorization approach to the numerical similarity matrix.

Feature learning on customer journey using categorical sequence data

A method and system to perform spatio-temporal prediction are described. The method includes obtaining, based on communication with one or more sources, multi-scale spatial datasets, each of the multi-scale spatial datasets providing a type of information at a corresponding granularity, at least two of the multi-scale spatial datasets providing at least two types of information at different corresponding granularities. The method also includes generating new features for each of the multi-scale spatial datasets, the new features being based on features of each of the multi-scale spatial datasets and spatial relationships between and within the multi-scale spatial datasets. The method further includes selecting, using the processor, features of interest from among the new features, training a predictive model based on the features of interest, and predicting an event based on the predictive model.

Auto-analyzing spatial relationships in multi-scale spatial datasets for spatio-temporal prediction

Electronic health records of a plurality of patients are received. A risk prediction model for a disease based on the electronic health records of the plurality of patients is created. An electronic health record of an original patient is received. A neighboring group of patients of the plurality of patients is identified, wherein the neighboring group of patients is two or more patients similar to the original patient. An ordering of the two or more patients of the neighboring group of patients is received, wherein the ordering of the two or more patients of the neighboring group of patients is based upon how similar each patient of the two or more patients is to the original patient. The risk prediction model is updated based on the ordering of the two or more patients of the neighboring group of patients.

Active patient risk prediction

Geo-defect repair modeling with location uncertainty is provided. A method includes logically dividing a railroad network into segments each of a specified length. The method also includes identifying, via a computer processor, geo-defects and approximated locations of the geo-defects occurring at each inspection run for each of the segments. The method also includes calculating, via the computer processor, a rate of increase in amplitude of each of the geo-defects for each of the segments between inspection runs, determining a correlation of the geo-defects between the inspection runs as a function of the approximated locations, and predicting a deterioration rate for each of the geo-defects based on the calculating.

Asset failure prediction with location uncertainty

A method, system and computer program product for detecting anomalies in a metering system. In one embodiment, data representing usage of a defined commodity are collected from meters, and the data collected over a given time period are analyzed to identify any of the meters showing at least one defined type of anomalous usage pattern. For each of the meters showing an anomalous usage patterns, an anomaly score is determined for the usage pattern shown, and the anomaly scores are used to rank the meters. In one embodiment, the collected data are analyzed to identify any of the meters showing one or more of a group of types of anomalous patterns including meter-no-reading, meter-by-pass, and meter-silting patterns. Embodiments of the invention utilize time series techniques and data analysis on meter reading data. Further, embodiments of the invention require no additional installation of equipment or sensors.

Anomaly detection using usage data for metering system

Methods and a system are provided that is performed by a computer server for inferring location context categories for a set of mobile users having at least two members. A method includes, for each mobile user in the set, obtaining at least one location context category therefor from publically available information responsive to uncertain mobile device location data. The method further includes applying multi-user collaborative machine learning to the at least one location context category for each mobile user in the set to infer a single refined location context category for each mobile user in the set.

Location context inference based on user mobile data with uncertainty

Techniques for real-time modeling temperature distributions based on streaming sensor data are provided. In one aspect, a method for creating a three-dimensional temperature distribution model for a room having a floor and a ceiling is provided. The method includes the following steps. A ceiling temperature distribution in the room is determined. A floor temperature distribution in the room is determined. An interpolation between the ceiling temperature distribution and the floor temperature distribution is used to obtain the three-dimensional temperature distribution model for the room.

Real-time modeling of heat distributions

Geo-defect repair modeling is provided. A method includes logically dividing a railroad network according to spatial and temporal dimensions with respect to historical data collected. The spatial dimensions include line segments of a specified length and the temporal dimensions include inspection run data for inspections performed for each of the line segments over a period of time. The method also includes creating a track deterioration model from the historical data, identifying geo-defects occurring at each inspection run from the track deterioration model, calculating a track deterioration condition from the track deterioration model by analyzing quantified changes in the geo-defects measured at each inspection run, and calculating a derailment risk based on track conditions determined from the inspection run data and the track deterioration condition. The method further includes determining a repair decision for each of the geo-defects based on the derailment risk and costs associated with previous comparable repairs.

Railway track geometry defect modeling for predicting deterioration, derailment risk, and optimal repair

Methods, systems and apparatus for determining a proposed cost for use of sensor resources and selecting a method of processing of sensed data are presented. Such a method includes the steps of: predicting a state of an environment, obtaining at least one criterion related to the use of the sensor resources comprising one or more data sensors, determining the proposed cost for use of the sensor resources, acquiring the sensed data from the one or more data sensors, determining a characteristic of the sensed data, and selecting the method of processing the sensed data according to the determined characteristic and the predicted state. The at least one criterion is based upon the predicted state. The determining of the proposed cost is based on the at least one criterion. One or more of steps are implemented on the processor device.

Determining cost and processing of sensed data

Growing community of inventors

Briarcliff Manor, NY, United States of America

Hongfei Li