System and method for resource adaptive classification of data streams
AssigneeInternational Business Machines Corporation
InventorsCharu C. Aggarwal, Philip Shi-lung Yu
Life of this patent
Estimating the term…
Abstract
A system and method for resource adaptive classification of data streams. Embodiments of systems and methods provide classifying data received in a computer, including discretizing the received data, constructing an intermediate data structure from said received data as training instances, performing subspace sampling on said received data as test instances and adaptively classifying said received data based on statistics of said subspace sampling.
Claims16
A system for classifying data received in a computer, the system comprising:
- a computer coupled to a memory for receiving the data in blocks smaller than said memory;
- and
- a process residing in said memory, the process having instructions to:
- construct an inverted index from said blocks of data;
- construct an inverted list for a range limit of training instances from said blocks of data, the range limit corresponding to a set of equi-depth ranges with an equal number of points in each range;
- sample subspaces on test instances from said blocks of data;
- and
- determine a class label based on a computation of an intersection of subspace samples, said intersection being between two or more of said inverted lists.
A method for classifying data received in a computer, the method comprising:
- receiving, in the computer, data in blocks smaller than a memory;
- constructing, in the computer, an inverted index from said blocks of data;
- constructing, in the computer, an inverted list for a range limit of training instances from said blocks of data, the range limit corresponding to a set of equi-depth ranges with an equal number of points in each range;
- sampling subspaces on test instances from said blocks of data;
- and
- determining, in the computer, a class label based on a computation of an intersection of subspace samples, said intersection being between two or more of said inverted lists.
A system comprising a server connected to a client computer, the system for performing a method, the method comprising:
- receiving streamed data in the server;
- constructing an inverted index of training instances from the streamed data;
- sampling subspaces from test instances received from the streamed data;
- and
- adaptively classifying the streamed data based on intersection points of said sampled subspaces.
A storage medium with machine-readable computer program code for classifying data, the storage medium including instructions for causing said computer to implement a method, comprising:
- constructing an inverted index structure including training instances from the data;
- and
- utilizing said inverted index structure for resource adaptive classification of the data.
A method for classifying data, the method, comprising:
- constructing an inverted index structure including training instances from data;
- and
- utilizing said inverted index structure for resource adaptive classification of the data.
Dependencies are inferred from the claim text.