Method and device for extracting feature value of time series data
Provided are a method and a device for extracting a feature value of time series data, the method including receiving time series data, segmenting the time series data into one or more segments according to preset segment types, identifying a first feature group mapped to a segment type of a first segment among the one or more segments, and extracting a value of at least one item included in the first feature group from the first segment as a feature value.
1. A method for extracting a feature value of time series data performed by a feature value extraction device of time series data, the method comprising:
acquiring, by the feature value extraction device, first time series data for a process in a semiconductor process;
segmenting, by the feature value extraction device, the first time series data into one or more segments according to preset segment types;
if a length of a first time section having a same value in the first time series data is less than a first threshold value:
compressing data of the first time section into data at a first time point included in the first time section; and
determining, in the compressed data, a second time section as a segment of one of the preset segment types based on a slope of the second time section including the first time point;
identifying, by the feature value extraction device, a first feature group of a plurality of preset feature groups mapped to a segment type of a first segment among the one or more segments;
extracting, by the feature value extraction device, a value of at least one item included in the first feature group as a first feature value of the first segment;
acquiring, by the feature value extraction device, a second time series data for said process in the semiconductor process, which is different from the first time series data;
synchronizing, to account for a time shift between the first time series data and the second time series data and by the feature value extraction device, the first segment of the first time series data and a second segment among one or more segments of the second time series data;
extracting, by the feature value extraction device, a value of the at least one item included in the first feature group as a second feature value of the second segment;
managing, by the feature value extraction device, the first feature value corresponding to the first segment and the second feature value corresponding to the second segment; and
determining, by the feature value extraction device, an anomaly in the semiconductor process corresponding to the second time series data based on the comparison of the first feature value and the second feature value.
2. The method of claim 1 , further comprising restoring the compressed data based on at least one of information on the first time point and information on the length of the first time section,
wherein a second segment included in the restored data corresponds to the second time section of the compressed data.
3. The method of claim 2 , further comprising:
adjusting a section of the second segment based on at least one of a second feature group mapped to a segment type of the second segment.
4. The method of claim 3 , further comprising:
adjusting the section of the second segment based on a slope of the second segment.
5. The method of claim 1 , further comprising:
if a number of the segmented segments is less than a second threshold value, adjusting a compression ratio of one or more segments included in the compressed data.
6. The method of claim 1 , wherein items included in the first feature group include at least one of a slope, a starting time point, an ending time point, a length, a standard deviation, a maximum value, a minimum value, a convexity, a residual, a kurtosis, a skewness, an amplitude, and a frequency, of a segment.
7. The method of claim 1 , wherein the preset segment types comprise at least one of a constant type, a spike type, a piecewise-constant type, a steady-state type, an up-ramp type, and a down-ramp type.
8. The method of claim 1 , wherein the segment type and a standard for determining the segment type are determined according to a type of the first time series data.
9. A device for extracting a feature value of time series data, the device comprising:
a memory configured to store at least one instruction; and
a processor,
wherein the processor, by executing the at least one instruction, is configured to:
acquire first time series data for a process in a semiconductor process;
segment the first time series data into one or more segments according to preset segment types;
if a length of a first time section having a same value in the first time series data is less than a first threshold value:
compress data of the first time section into data at a first time point included in the first time section; and
determine, in the compressed data, a second time section as a segment of one of the preset segment types based on a slope of the second time section including the first time point;
identify a first feature group of a plurality of preset feature groups mapped to a segment type of a first segment among the one or more segments;
extract a value of at least one item included in the first feature group as a first feature value of the first segment;
acquire a second time series data for said process in the semiconductor process, which is different from the first time series data;
synchronize, to account for a time shift between the first time series data and the second time series data, the first segment of the first time series data and a second segment among one or more segments of the second time series data;
extract a value of the at least one item included in the first feature group as a second feature value of the second segment;
manage the first feature value corresponding to the first segment and the second feature value corresponding to the second segment; and
determine an anomaly in the semiconductor process corresponding to the second time series data based on the comparison of the first feature value and the second feature value.
10. The device of claim 9 , wherein the processor is further configured to: restore the compressed data based on at least one of information on the first time point and information on the length of the first time section,
wherein a second segment included in the restored data corresponds to the second time section of the compressed data.
11. The device of claim 10 , wherein the processor is further configured to:
adjust a section of the second segment based on at least one of a second feature group mapped to a segment type of the second segment.
12. The device of claim 11 , wherein the processor is further configured to:
adjust the section of the second segment based on a slope of the second segment.
13. The device of claim 9 , wherein the processor is further configured to:
if a number of the segmented segments is less than a second threshold value, adjust a compression ratio of one or more segments included in the compressed data.
14. The device of claim 9 , wherein items included in the first feature group include at least one of a slope, a starting time point, an ending time point, a length, a standard deviation, a maximum value, a minimum value, a convexity, a residual, a kurtosis, a skewness, an amplitude, and a frequency, of a segment.
15. A non-transitory computer-readable recording medium in which a program to execute a method for extracting a feature value of time series data in a computer is recorded, the method comprising:
acquiring first time series data for a process in a semiconductor process;
segmenting the first time series data into one or more segments according to preset segment types;
if a length of a first time section having a same value in the first time series data is less than a first threshold value:
compressing data of the first time section into data at a first time point included in the first time section; and
determining, in the compressed data, a second time section as a segment of one of the preset segment types based on a slope of the second time section including the first time point;
identifying a first feature group of a plurality of preset feature groups mapped to a segment type of a first segment among the one or more segments;
extracting a value of at least one item included in the first feature group as a first feature value of the first segment;
acquiring a second time series data for said process in the semiconductor process, which is different from the first time series data;
synchronizing, to account for a time shift between the first time series data and the second time series data, the first segment of the first time series data and a second segment among one or more segments of the second time series data;
extracting a value of the at least one item included in the first feature group as a second feature value of the second segment;
managing the first feature value corresponding to the first segment and the second feature value corresponding to the second segment; and
determining an anomaly in the semiconductor process corresponding to the second time series data based on the comparison of the first feature value and the second feature value.
16. The non-transitory computer-readable recording medium of claim 15 , the method further comprising restoring the compressed data based on at least one of information on the first time point and information on the length of the first time section, wherein a second segment included in the restored data corresponds to the second time section of the compressed data.
17. The non-transitory computer-readable recording medium of claim 16 , the method further comprising:
adjusting a section of the second segment based on at least one of a second feature group mapped to a segment type of the second segment.
18. The non-transitory computer-readable recording medium of claim 17 , the method further comprising:
adjusting the section of the second segment based on a slope of the second segment.
19. The non-transitory computer-readable recording medium of claim 15 , the method further comprising:
if a number of the segmented segments is less than a second threshold value, adjusting a compression ratio of one or more segments included in the compressed data.
20. The non-transitory computer-readable recording medium of claim 15 , wherein items included in the first feature group include at least one of a slope, a starting time point, an ending time point, a length, a standard deviation, a maximum value, a minimum value, a convexity, a residual, a kurtosis, a skewness, an amplitude, and a frequency, of a segment.