ASSESSING ACCURACY OF TRAINED PREDICTIVE MODELS

• US 20130346351A1
• Filed: 08/20/2013
• Published: 12/26/2013
• Est. Priority Date: 05/04/2011
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Lin, Wei-Hao, Green, Travis H. K., Kaplow, Robert, Mann, Gideon S., Fu, Gang
• Granted Patent: US 9,239,986 B2
• Time in Patent Office: Days
• US Class Current: 706/21
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/08 Learning methods

Abstract

A system includes a computer(s) coupled to a data storage device(s) that stores a training data repository and a predictive model repository. The training data repository includes retained data samples from initial training data and from previously received data sets. The predictive model repository includes at least one updateable trained predictive model that was trained with the initial training data and retrained with the previously received data sets. A new data set is received. A richness score is assigned to each of the data samples in the set and to the retained data samples that indicates how information rich a data sample is for determining accuracy of the trained predictive model. A set of test data is selected based on ranking by richness score the retained data samples and the new data set. The trained predictive model is accuracy tested using the test data and an accuracy score determined.

Claims

1-21. -21. (canceled)

22. A computer-implemented method comprising:

receiving a first data set of data samples, each data sample comprising input data and corresponding output data, wherein the first data set is new relative to a retained data set of data samples, where each data sample in the retained data set comprising input data and corresponding output data, and where the retained data set was used in training predictive models in a repository of predictive models;

determining a richness score for each of the data samples included in the first data set and to each of the data samples in the retained data set, wherein the richness score for a particular data sample indicates how information rich the particular data sample is, relative to other data samples in the set of retained data samples and the first data set, for determining an accuracy of a trained predictive model, and where the richness score for a particular data sample is based, at least in part, on how redundant the particular data sample is of other, different data samples;

ranking the data samples included in the first data set and the set of retained data samples based on the assigned richness scores; and

selecting a first set of test data from the data samples included in the first data set and the set of retained data samples based on the ranking.

23. The computer-implemented method of claim 22, further comprising:

testing how accurate each of the trained predictive models in the repository is in determining predictive output data for given input data using the first set of test data and determining respective accuracy scores for each of the trained predictive model based on the testing; and

selecting a first trained predictive model from the repository based on the accuracy scores and providing access to the first trained predictive model to a client computing system for generating predictive output data based on input data received from the client computing system.

24. The computer-implemented method of claim 22, wherein determining a richness score for each of the data samples comprises determining the richness score based on a ratio of how many data samples have similar input data but different output data than the data sample to how many data samples have similar input data and similar or different output data than the data sample.

25. The computer-implemented method of claim 22, wherein determining a richness score for each of the data samples comprises:

determining a cluster to which the data sample belongs;

in response to determining the data samples to the cluster exhibit a high degree of similarity, determining a relatively high richness score for the data sample and at least one other data sample that belongs to the cluster relative to other richness scores for other data samples that belong to the cluster.

26. The computer-implemented method of claim 22, wherein determining a richness score for each of the data samples comprises:

determining a cluster to which the data sample belongs;

determining a time value for each data sample that belongs to the cluster, the time value indicative of how recent a data sample was received; and

determining a richness score that is proportional to the recency of the time value.

27. The computer-implemented method of claim 22, wherein determining a richness score for each of the data samples comprises:

determining a cluster to which the data sample belongs;

determining a relative position of the data sample from an edge of the cluster; and

determining a richness score that is inversely proportional to a distance of the relative position of the data sample from the edge of the cluster.

28. The computer-implemented method of claim 22, wherein determining a richness score for each of the data samples comprises:

determining a cluster to which the data sample belongs;

determining a relative position of the data sample from an edge of the cluster; and

determining a richness score that is inversely proportional to a distance of the relative position of the data sample from the edge of the cluster.

29. The computer-implemented method of claim 22, wherein determining a richness score for each of the data samples comprises:

determining a cluster to which the data sample belongs;

determining first richness scores for the data sample and other data samples that belong to the cluster;

determining second richness scores for the other data samples that belong to the cluster when the data sample is removed from the cluster;

setting the richness score of the data sample to a null value when a sum of second richness scores for the other data samples are greater than a sum of the first richness scores for the other data samples.

30. A computer-implemented system comprising:

one or more computers; and

one or more data storage devices coupled to the one or more computers, storing instructions that, when executed by the one or more computers, cause the one or more computers to perform operations comprising;

receiving a first data set of data samples, each data sample comprising input data and corresponding output data, wherein the first data set is new relative to a retained data set of data samples, where each data sample in the retained data set comprising input data and corresponding output data, and where the retained data set was used in training predictive models in a repository of predictive models;

determining a richness score for each of the data samples included in the first data set and to each of the data samples in the retained data set, wherein the richness score for a particular data sample indicates how information rich the particular data sample is, relative to other data samples in the set of retained data samples and the first data set, for determining an accuracy of a trained predictive model, and where the richness score for a particular data sample is based, at least in part, on how redundant the particular data sample is of other, different data samples;

ranking the data samples included in the first data set and the set of retained data samples based on the assigned richness scores; and

selecting a first set of test data from the data samples included in the first data set and the set of retained data samples based on the ranking.

31. The computer-implemented system of claim 30, the instructions causing the one or more computers to perform operations further comprising:

testing how accurate each of the trained predictive models in the repository is in determining predictive output data for given input data using the first set of test data and determining respective accuracy scores for each of the trained predictive model based on the testing; and

selecting a first trained predictive model from the repository based on the accuracy scores and providing access to the first trained predictive model to a client computing system for generating predictive output data based on input data received from the client computing system.

32. The computer-implemented system of claim 30, wherein determining a richness score for each of the data samples comprises determining a richness score for each of the data samples comprises determining the richness score based on a ratio of how many data samples have similar input data but different output data than the data sample to how many data samples have similar input data and similar or different output data than the data sample.

33. The computer-implemented system of claim 30, wherein determining a richness score for each of the data samples comprises:

determining a cluster to which the data sample belongs;

in response to determining the data samples to the cluster exhibit a high degree of similarity, determining a relatively high richness score for the data sample and at least one other data sample that belongs to the cluster relative to other richness scores for other data samples that belong to the cluster.

34. The computer-implemented system of claim 30, wherein determining a richness score for each of the data samples comprises:

determining a cluster to which the data sample belongs;

determining a time value for each data sample that belongs to the cluster, the time value indicative of how recent a data sample was received; and

determining a richness score that is proportional to the recency of the time value.

35. The computer-implemented system of claim 30, wherein determining a richness score for each of the data samples comprises:

determining a cluster to which the data sample belongs;

determining a relative position of the data sample from an edge of the cluster; and

determining a richness score that is inversely proportional to a distance of the relative position of the data sample from the edge of the cluster.

36. The computer-implemented system of claim 30, wherein determining a richness score for each of the data samples comprises:

determining a cluster to which the data sample belongs;

determining a relative position of the data sample from an edge of the cluster; and

determining a richness score that is inversely proportional to a distance of the relative position of the data sample from the edge of the cluster.

37. The computer-implemented system of claim 30 wherein determining a richness score for each of the data samples comprises:

determining a cluster to which the data sample belongs;

determining first richness scores for the data sample and other data samples that belong to the cluster;

determining second richness scores for the other data samples that belong to the cluster when the data sample is removed from the cluster;

setting the richness score of the data sample to a null value when a sum of second richness scores for the other data samples are greater than a sum of the first richness scores for the other data samples.

38. A computer-readable storage device encoded with a computer program product, the computer program product comprising instructions that when executed on one or more computers cause the one or more computers to perform operations comprising:

receiving a first data set of data samples, each data sample comprising input data and corresponding output data, wherein the first data set is new relative to a retained data set of data samples, where each data sample in the retained data set comprising input data and corresponding output data, and where the retained data set was used in training predictive models in a repository of predictive models;

determining a richness score for each of the data samples included in the first data set and to each of the data samples in the retained data set, wherein the richness score for a particular data sample indicates how information rich the particular data sample is, relative to other data samples in the set of retained data samples and the first data set, for determining an accuracy of a trained predictive model, and where the richness score for a particular data sample is based, at least in part, on how redundant the particular data sample is of other, different data samples;

ranking the data samples included in the first data set and the set of retained data samples based on the assigned richness scores; and

selecting a first set of test data from the data samples included in the first data set and the set of retained data samples based on the ranking.

39. The computer-readable storage device of claim 38, the instructions causing the one or more computers to perform operations further comprising:

testing how accurate each of the trained predictive models in the repository is in determining predictive output data for given input data using the first set of test data and determining respective accuracy scores for each of the trained predictive model based on the testing; and

selecting a first trained predictive model from the repository based on the accuracy scores and providing access to the first trained predictive model to a client computing system for generating predictive output data based on input data received from the client computing system.

40. The computer-readable storage device of claim 38, wherein determining a richness score for each of the data samples comprises determining a richness score for each of the data samples comprises determining the richness score based on a ratio of how many data samples have similar input data but different output data than the data sample to how many data samples have similar input data and similar or different output data than the data sample.

41. The computer-readable storage device of claim 38, wherein determining a richness score for each of the data samples comprises:

determining a cluster to which the data sample belongs;

in response to determining the data samples to the cluster exhibit a high degree of similarity, determining a relatively high richness score for the data sample and at least one other data sample that belongs to the cluster relative to other richness scores for other data samples that belong to the cluster.

42. The computer-readable storage device of claim 38, wherein determining a richness score for each of the data samples comprises:

determining a cluster to which the data sample belongs;

determining first richness scores for the data sample and other data samples that belong to the cluster;

determining second richness scores for the other data samples that belong to the cluster when the data sample is removed from the cluster;

setting the richness score of the data sample to a null value when a sum of second richness scores for the other data samples are greater than a sum of the first richness scores for the other data samples.