AUTOMATIC TIME SERIES EXPLORATION FOR BUSINESS INTELLIGENCE ANALYTICS

• US 20160342910A1
• Filed: 05/13/2016
• Published: 11/24/2016
• Est. Priority Date: 05/18/2015
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Chu, Yea Jane, Han, Sier, Shyr, Jing-Yun
• Granted Patent: US 10,453,008 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06Q 10/063 Operations research, analysis or management
• CPC Class Codes: G06T 11/206 Drawing of charts or graphs

Abstract

Techniques are described for generating characterizations of time series data. In one example, a method includes extracting a trend-cycle component, a seasonal component, and an irregular component from a time series of data. The method further includes performing one or more pattern analyses on the trend-cycle component, the seasonal component, and the irregular component. The method further includes, for each pattern analysis of the one or more pattern analyses, performing a comparison of an analytic result of the respective pattern analysis to a selected significance threshold for the respective pattern analysis to determine if the analytic result passes the significance threshold for the respective pattern analysis. The method further includes generating an output for each of the analytic results that pass the significance threshold for the respective pattern analysis.

Claims

1. A method for generating characterizations of time series data, the method comprising:

extracting a trend-cycle component, a seasonal component, and an irregular component from a time series of data;

performing one or more pattern analyses on the trend-cycle component, the seasonal component, and the irregular component;

for each pattern analysis of the one or more pattern analyses, performing a comparison of an analytic result of the respective pattern analysis to a selected significance threshold for the respective pattern analysis to determine if the analytic result passes the significance threshold for the respective pattern analysis; and

generating an output for each of the analytic results that pass the significance threshold for the respective pattern analysis.

2. The method of claim 1,

wherein performing the one or more pattern analyses on the trend-cycle component comprises performing an overall trend analysis using a time correlation statistic, and

wherein performing the comparison of the analytic result of the respective pattern analysis to the selected significance threshold for the respective pattern analysis comprises performing a comparison of the result of the time correlation statistic to a selected time correlation significance threshold.

3. The method of claim 1,

wherein performing the one or more pattern analyses on the trend-cycle component comprises performing a turning point detection analysis, and

wherein performing the comparison of the analytic result of the respective pattern analysis to the selected significance threshold for the respective pattern analysis comprises performing a comparison of the result of the turning point detection analysis to a selected turning point significance threshold.

4. The method of claim 1,

wherein performing the one or more pattern analyses on the seasonal component comprises performing a seasonal pattern significance test, and

wherein performing the comparison of the analytic result of the respective pattern analysis to the selected significance threshold for the respective pattern analysis comprises performing a comparison of a result of a p-value of the seasonal pattern significance test to an alpha significance threshold.

5. The method of claim 1,

wherein performing the one or more pattern analyses on the seasonal component comprises determining an effect size value to indicate a strength of the seasonal component, and

wherein performing the comparison of the analytic result of the respective pattern analysis to the selected significance threshold for the respective pattern analysis comprises performing a comparison of the result of the effect size value to one or more selected effect size significance thresholds.

6. The method of claim 1,

wherein performing the one or more pattern analyses on the seasonal component comprises performing an unusual season detection analysis, and

wherein performing the comparison of the analytic result of the respective pattern analysis to the selected significance threshold for the respective pattern analysis comprises performing a comparison of a result of z-scores of seasons to a selected positive or negative z-score significance threshold.

7. The method of claim 1,

wherein performing the one or more pattern analyses on the irregular component comprises identifying a time of the irregular component and performing an outlier detection analysis of the irregular component, and

wherein performing the comparison of the analytic result of the respective pattern analysis to the selected significance threshold for the respective pattern analysis comprises performing a comparison of a result of z-score values from the outlier detection analysis to a selected z-score significance threshold.

8. The method of claim 1,

wherein performing the one or more pattern analyses on the irregular component comprises performing a large variance interval detection analysis, and

wherein performing the comparison of the analytic result of the respective pattern analysis to the selected significance threshold for the respective pattern analysis comprises performing a comparison of the result of a p-value of a large variance interval test statistic from the large variance interval detection analysis to a selected alpha significance threshold.

9. The method of claim 1,

wherein performing the one or more pattern analyses on the irregular component comprises performing an autocorrelation function analysis, and

wherein performing the comparison of the analytic result of the respective pattern analysis to the selected significance threshold for the respective pattern analysis comprises performing a comparison of the result of the autocorrelation value to a selected confidence level significance threshold.

10. The method of claim 1, further comprising:

distributing the time series of data into different mappers, and

performing one or more of the pattern analyses in parallel.