Storage and analysis of time-series data based on data context and patterns

1 . A method for compressing data, comprising:

at a computer system having one or more processors and memory:

obtaining a plurality of time-series data sets, each of which includes a temporally indexed sequence of data items;

identifying a first data pattern that is shared by the plurality of time-series data sets, the first data pattern including a first subset of successive data items;

representing the first data pattern with a first pattern indicator;

replacing the first subset of successive data items corresponding to the first data pattern with the first pattern indicator in each of the plurality of time-series data sets;

applying a pattern analysis model to process the first data pattern that is shared by the plurality of time-series data sets and generate a pattern control output;

modifying an operation of a machine, external to and distinct from the computer system, based on the control pattern output, wherein the operation of the machine includes a programmable control arm; and

storing the first data pattern and the plurality of time-series data sets in a database, each of the plurality of time-series data sets including the first pattern indicator.

2 . The method of claim 1 , further comprising:

obtaining a new time-series data set;

determining that the new time-series data set includes a portion that corresponds to the first data pattern; and

replacing the portion of the new time-series data set corresponding to the first data pattern with the first pattern indicator.

3 . The method of claim 1 , wherein the first subset of successive data items corresponding to the first data pattern has a respective temporal location in each of the plurality of time-series data sets, and respective temporal locations of the first data pattern in the plurality of time-series data sets are aligned with one another.

4 . The method of claim 1 , further comprising:

identifying a second data pattern that is shared by a subset of two or more time-series data sets of the plurality of time-series data sets, the second data pattern distinct from the first data pattern and including a second subset of successive data items, the subset of two or more time-series data sets including less than all of the plurality of time-series data sets.

5 . The method of claim 4 , further comprising:

applying a second pattern indicator in place of the second subset of successive data items corresponding to the second data pattern in each of the subset of two or more time-series data sets, wherein each of the second subset of successive time-series data sets is stored with the second pattern indicator in place of the second data pattern.

6 . The method of claim 1 , wherein the plurality of time-series data sets includes a first time-series data set including the first subset of successive data items and a remainder set of data items, the method further comprising:

representing the remainder set of data items with one or more pattern indicators, wherein the one or more pattern indicators include a third pattern indicator, and the third pattern indicator corresponds to a third subset of one or more successive data items that appears at a corresponding temporal location within the first time-series data set.

7 . The method of claim 6 , further comprising:

adding the third subset of one or more successive data items, corresponding to a third data pattern, into a corpus of training data items to be used by a machine learning system to generate an anomaly detection model for detecting anomalies in a physical process performed by the machine.

8 . The method of claim 6 , further comprising:

modifying at least one data item in the third subset of one or more successive data items, corresponding to a third data pattern, to generate synthetic training data having a modified subset of one or more successive data items; and

adding the synthetic training data into a corpus of training data items to be used by a machine learning system to generate an anomaly detection model for detecting anomalies in a physical process performed by the machine.

9 . The method of claim 1 , wherein each of the first subset of successive data items of the first data pattern has a respective data value, the method further comprising, for each of the plurality of time-series data sets:

determining that the respective time-series data set includes a respective set of successive data items, each data item of which is equal to the respective data value of a corresponding data item of the first subset of successive data items of the first data pattern.

10 . The method of claim 1 , wherein each of the first subset of successive data items of the first data pattern has a respective value tolerance range, and the method further comprises, for each of the plurality of time-series data sets:

determining that the respective time-series data set includes a respective set of successive data items, each data item of which is within the respective value tolerance range of a corresponding data item of the first subset of successive data items of the first data pattern.

11 . The method of claim 1 , wherein each of the first subset of successive data items of the first data pattern has a respective data change rate, further comprising, for each of the plurality of time-series data sets:

determining that the respective time-series data set includes a set of successive data items, each data item of which is equal to, or within a respective change tolerance range of, the respective data change rate of a corresponding data item of the first subset of successive data items of the first data pattern.

12 . The method of claim 1 , wherein storing the first data pattern further comprises:

quantizing each data item of the first subset of successive data items of the first data pattern to generate a quantized first data pattern; and

storing the quantized first data pattern.

13 . A computer system, comprising:

one or more processors; and

memory storing one or more programs for execution by the one or more processors, the one or more programs further comprising instructions for:

obtaining a plurality of time-series data sets, each of which includes a temporally indexed sequence of data items;

identifying a first data pattern that is shared by the plurality of time-series data sets, the first data pattern including a first subset of successive data items;

representing the first data pattern with a first pattern indicator;

replacing the first subset of successive data items corresponding to the first data pattern with the first pattern indicator in each of the plurality of time-series data sets;

applying a pattern analysis model to process the first data pattern that is shared by the plurality of time-series data sets and generate a pattern control output;

modifying an operation of a machine, external to and distinct from the computer system, based on the control pattern output, wherein the operation of the machine includes a programmable control arm; and

storing the first data pattern and the plurality of time-series data sets in a database, each of the plurality of time-series data sets including the first pattern indicator.

14 . The computer system of claim 13 , wherein:

the plurality of time-series data sets are obtained from an electronic device that is part of the machine; and

obtaining the plurality of time-series data sets further comprises obtaining each of the plurality of time-series data sets from the electronic device while the machine is implementing a respective robotic operation according to a predefined operation configuration;

wherein modifying the operation of the machine includes adjusting the predefined operation configuration of the machine based on the pattern control output.

15 . The computer system of claim 13 , the one or more programs further comprising instructions for:

modifying at least one data item in the first subset of one or more successive data items, corresponding to the first pattern indicator, to generate a modified first subset of one or more successive data items; and

adding the modified first subset of one or more successive data items into a corpus of training data items.

16 . The computer system of claim 13 , the one or more programs further comprising instructions for:

obtaining a new time-series data set;

determining that the new time-series data set includes a portion that corresponds to the first data pattern; and

replacing the portion of the new time-series data set corresponding to the first data pattern with the first pattern indicator.

17 . A non-transitory computer-readable storage medium, storing one or more programs for execution by one or more processors, the one or more programs further comprising instructions for:

obtaining a plurality of time-series data sets, each of which includes a temporally indexed sequence of data items;

identifying a first data pattern that is shared by the plurality of time-series data sets, the first data pattern including a first subset of successive data items;

representing the first data pattern with a first pattern indicator;

replacing the first subset of successive data items corresponding to the first data pattern with the first pattern indicator in each of the plurality of time-series data sets;

applying a pattern analysis model to process the first data pattern that is shared by the plurality of time-series data sets and generate a pattern control output;

modifying an operation of a machine based on the control pattern output, wherein the operation of the machine includes a programmable control arm; and

storing the first data pattern and the plurality of time-series data sets in a database, each of the plurality of time-series data sets including the first pattern indicator.

18 . The non-transitory computer-readable storage medium of claim 17 , wherein identifying the first data pattern that is shared by the plurality of time-series data sets further comprises:

applying dynamic time warping to identify the first data pattern at a first temporal location in a first time-series data set and at a second temporal location in a second time-series data set, respectively, wherein the second temporal location is distinct from the first temporal location.

19 . The non-transitory computer-readable storage medium of claim 17 , wherein identifying the first data pattern that is shared by the plurality of time-series data sets further comprises:

determining that the first data pattern corresponding to a first data segment in a first time-series data set, wherein a first length of the first data segment is equal to a length of the first data pattern; and

determining that the first data pattern corresponding to a second data segment in a second time-series data set, including stretching a length of the second data segment to match the length of the first data pattern and adding one or more data items to supplement the second data segment of the second time-series data set to match the first data pattern.

20 . The non-transitory computer-readable storage medium of claim 17 , wherein identifying the first data pattern that is shared by the plurality of time-series data sets further comprises:

determining that the first data pattern corresponding to a first data segment in a first time-series data set, wherein a first length of the first data segment is equal to a length of the first data pattern; and

determining that the first data pattern corresponding to a second data segment in a second time-series data set, including scaling down a length of the second data segment to match the length of the first data pattern and removing one or more data items to match the second data segment of the second time-series data set to the first data pattern.