Flexible algorithm for time dimension ranking
The present disclosure involves systems, software, and computer implemented methods for ranking time dimensions. One example method includes receiving a request for an insight analysis for a dataset that includes a value dimension and a set of multiple date dimensions. Each date dimension is converted into a time series and a value quality factor is determined for each time series that represents a level of data quality for the time series. A time series informative factor is determined for each time series that represents how informative the time series is within a specified time window. An insight score is determined, for each time dimension, based on the determined value quality factors and the determined time series informative factors. The insight score for the time dimension is provided, for at least some of the time dimensions.
1. A computer-implemented method comprising:
receiving a request for an insight analysis for a dataset, wherein the dataset includes a value dimension and a set of multiple date dimensions, wherein at least some of the date dimensions have missing values over a time range of dates included in the multiple date dimensions, and wherein the request includes a specified time window within the time range;
generating multiple times series by converting each date dimension of the multiple date dimensions into a respective time series;
determining, for each respective time series of the multiple time series, a value quality factor that represents a level of data quality for the respective time series;
determining, for each respective time series of the multiple time series, a time series informative factor that represents how informative the respective time series is within the specified time window;
determining, based on the determined value quality factors and the determined time series informative factors, an insight score for each time series of the multiple time series, wherein a respective insight score for a respective time series is based on both the value quality factor for the respective time series that represents the level of quality for the respective time series and the time series informative factor for the respective time series, and wherein the respective insight score for the respective time series that represents both how informative the respective time series is within the specified time window and a respective likelihood that the respective time series exhibits unstable behavior as compared to other time series of the multiple time series;
determining, from among the insight scores of the multiple time series, a set of highest-ranked insight scores for a set of highest ranked time series; and
providing at least some of the highest-ranked time series and the insight scores for the highest-ranked time series to a machine learning system.
2. The method of claim 1 , wherein a higher insight score for a time series represents a higher level of insight provided by the time series.
3. The method of claim 1 , wherein determining the insight score for a given time series comprises multiplying the value quality factor for the time series by the times series informative factor for the time series.
4. The method of claim 1 , wherein the value quality factor for a time series represents a number of missing values in the time series and distribution of time points in the time series.
5. The method of claim 1 , wherein the time series informative factor for a time series is based on changes in the value dimension for dates in the specified time window.
6. The method of claim 5 , wherein the time series informative factor for a time series is based on changes in the specified time window relative to changes in the time series.
7. The method of claim 1 , wherein determining the time series informative factor for a time series comprises:
generating time segmentations for time points in the specified time window;
determining metrics for the generated time segmentations; and
comparing the metrics for the generated time segmentations to corresponding metrics for all time points in the time series.
8. The method of claim 7 , wherein determining the time series informative factor for a time series comprises using a decaying weight vector to assign different weights to different time segmentations.
9. A system comprising:
one or more computers; and
a computer-readable medium coupled to the one or more computers having instructions stored thereon which, when executed by the one or more computers, cause the one or more computers to perform operations comprising:
receiving a request for an insight analysis for a dataset, wherein the dataset includes a value dimension and a set of multiple date dimensions, wherein at least some of the date dimensions have missing values over a time range of dates included in the multiple date dimensions, and wherein the request includes a specified time window within the time range;
generating multiple times series by converting each date dimension of the multiple date dimensions into a respective time series;
determining, for each respective time series of the multiple time series, a value quality factor that represents a level of data quality for the respective time series;
determining, for each respective time series of the multiple time series, a time series informative factor that represents how informative the respective time series is within the specified time window;
determining, based on the determined value quality factors and the determined time series informative factors, an insight score for each time series of the multiple time series, wherein a respective insight score for a respective time series is based on both the value quality factor for the respective time series that represents the level of quality for the respective time series and the time series informative factor for the respective time series, and wherein the respective insight score for the respective time series represents both how informative the respective time series is within the specified time window and a respective likelihood that the respective time series exhibits unstable behavior as compared to other time series of the multiple time series;
determining, from among the insight scores of the multiple time series, a set of highest-ranked insight scores for a set of highest ranked time series; and
providing at least some of the highest-ranked time series and the insight scores for the highest-ranked time series to a machine learning system.
10. The system of claim 9 , wherein a higher insight score for a time series represents a higher level of insight provided by the time series.
11. The system of claim 9 , wherein determining the insight score for a given time series comprises multiplying the value quality factor for the time series by the times series informative factor for the time series.
12. A computer program product encoded on a non-transitory storage medium, the product comprising non-transitory, computer readable instructions for causing one or more processors to perform operations comprising:
receiving a request for an insight analysis for a dataset, wherein the dataset includes a value dimension and a set of multiple date dimensions, wherein at least some of the date dimensions have missing values over a time range of dates included in the multiple date dimensions, and wherein the request includes a specified time window within the time range;
generating multiple times series by converting each date dimension of the multiple date dimensions into a respective time series;
determining, for each respective time series of the multiple time series, a value quality factor that represents a level of data quality for the respective time series;
determining, for each respective time series of the multiple time series, a time series informative factor that represents how informative the respective time series is within the specified time window;
determining, based on the determined value quality factors and the determined time series informative factors, an insight score for each time series of the multiple time series, wherein a respective insight score for a respective time series is based on both the value quality factor for the respective time series that represents the level of quality for the respective time series and the time series informative factor for the respective time series, and wherein the respective insight score for the respective time series represents both how informative the respective time series is within the specified time window and a respective likelihood that the respective time series exhibits unstable behavior as compared to other time series of the multiple time series;
determining, from among the insight scores of the multiple time series, a set of highest-ranked insight scores for a set of highest ranked time series; and
providing at least some of the highest-ranked time series and the insight scores for the highest-ranked time series to a machine learning system.
13. The computer program product of claim 12 , wherein a higher insight score for a time series represents a higher level of insight provided by the time series.
14. The computer program product of claim 12 , wherein determining the insight score for a given time series comprises multiplying the value quality factor for the time series by the times series informative factor for the time series.