Time-series forecasting based on detected downtime
Provided is a system and method which generates a composite machine learning model that can filter downtime data from a time-series data signal and perform a prediction on remaining time-series data. In one example, the method may include detecting a pattern of downtime data within a time-series data signal, removing a subset of data from the time-series data based on the detected pattern of downtime and building a machine learning model to make predictions based on remaining data in the time-series data, generating segregation instructions configured to remove downtime data from a time-series data signal of a same type and to predict zero on future dates matching the downtime segregation codes, and building a composite machine learning model that includes the trained machine learning model and the segregation instructions for filtering data that is input to the trained machine learning models.
1 . A computing system comprising:
a memory configured to store time-series data of a time-series data signal; and
a processor coupled to the memory and configured to:
detect one or more patterns of downtime within the time-series data signal based on null values within the time-series data;
remove a subset of data from the time-series data based on the one or more detected patterns of downtime;
build a machine learning model to make predictions based on remaining data in the time-series data;
generate segregation instructions configured to remove downtime data from a time-series data signal of a same type, the segregation instructions including encoded computer instructions to filter downtime data from time series data based on a combination of a frequency of occurrence value of a recurring pattern of downtime and a time period value at which the recurring pattern of downtime occurs;
build a composite machine learning model that includes the machine learning model and the segregation instructions for filtering data that is input to the machine learning model; and
store the composite model in the memory.
2 . The computing system of claim 1 , wherein the processor is further configured to deploy the composite model including the machine learning model and the segregation instructions within a productive environment.
3 . The computing system of claim 1 , wherein the processor is further configured to execute the machine learning model on new a new time-series data signal and output a prediction based on the new time-series data signal, wherein the processor automatically removes a subset of data from the new time-series data signal based on the segregation instructions prior to generating the prediction and enforces a zero value for the subset of data in the output prediction.
4 . The computing system of claim 1 , wherein the processor is configured to apply a plurality of different granularity values when detecting a pattern of downtime and select a granularity value from among the plurality of different granularity values to assign the pattern of downtime.
5 . The computing system of claim 4 , wherein the plurality of different granularity values comprise at least two of a daily value, a weekly value, a monthly value, and a quarterly value.
6 . The computing system of claim 1 , wherein the processor is configured to divide the time-series data signal into two subsets including a first subset with a fluctuating time-series signal value and a second subset with a null time-series signal value, and train the machine learning model using the first subset but not the second subset.
7 . The computing system of claim 1 , wherein the processor is further configured to receive downtime inputs via a user interface and generate additional segregation instructions for removing data from the time-series data signal of a same type based on the received downtime inputs.
8 . A method comprising:
detecting one or more patterns of downtime within a time-series data signal based on null values within values of time-series data corresponding to the time-series data signal;
removing a subset of data from the time-series data based on the one or more detected patterns of downtime and building a machine learning model to make predictions based on remaining data in the time-series data;
generating segregation instructions configured to remove downtime data from a time-series data signal of a same type, the segregation instructions including encoded computer instructions to filter downtime data from time series data based on a combination of a frequency of occurrence value of a recurring pattern of downtime and a time period value at which the recurring pattern of downtime occurs;
building a composite machine learning model that includes the machine learning model and the segregation instructions for filtering data that is input to the machine learning model; and
storing the composite model in memory.
9 . The method of claim 8 , wherein the method further comprises deploying the composite model including the machine learning model and the segregation instructions within a productive environment.
10 . The method of claim 8 , wherein the method further comprises executing the machine learning model on new a new time-series data signal and outputting a prediction based on the new time-series data signal, wherein the executing comprises automatically removing a subset of data from the new time-series data signal based on the segregation instructions prior to generating the prediction and enforcing a zero value for the subset of data in the output prediction.
11 . The method of claim 8 , wherein the detecting comprises applying a plurality of different granularity values when detecting a pattern of downtime and selecting a granularity value from among the plurality of different granularity values to assign the pattern of downtime.
12 . The method of claim 11 , wherein the plurality of different granularity values comprise at least two of a daily value, a weekly value, a monthly value, and a quarterly value.
13 . The method of claim 8 , wherein the removing comprises dividing the time-series data signal into two subsets including a first subset with a fluctuating time-series signal value and a second subset with a null time-series signal value, and training the machine learning model using the first subset but not the second subset.
14 . The method of claim 8 , wherein the method further comprises receiving downtime inputs via a user interface and generating additional segregation instructions for removing data from the time-series data signal of a same type based on the received downtime inputs.
15 . A non-transitory computer-readable medium comprising instructions which when executed by a processor cause a computer to perform a method comprising:
detecting one or more patterns of downtime within a time-series data signal based on null values within values of time-series data corresponding to the time-series data signal;
removing a subset of data from the time-series data based on the one or more detected patterns of downtime and building a machine learning model to make predictions based on remaining data in the time-series data;
generating segregation instructions configured to remove downtime data from a time- series data signal of a same type, the segregation instructions including encoded computer instructions to filter downtime data from time series data based on a combination of a frequency of occurrence value of a recurring pattern of downtime and a time period value at which the recurring pattern of downtime occurs;
building a composite machine learning model that includes the machine learning model and the segregation instructions for filtering data that is input to the machine learning model; and
storing the composite model in memory.
16 . The non-transitory computer-readable medium of claim 15 , wherein the method further comprises deploying the composite model including the machine learning model and the segregation instructions within a productive environment.
17 . The non-transitory computer-readable medium of claim 15 , wherein the method further comprises executing the machine learning model on new a new time-series data signal and outputting a prediction based on the new time-series data signal, wherein the executing comprises automatically removing a subset of data from the new time-series data signal based on the segregation instructions prior to generating the prediction and enforcing a zero value for the subset of data in the output prediction.
18 . The non-transitory computer-readable medium of claim 15 , wherein the detecting comprises applying a plurality of different granularity values when detecting a recurring pattern of downtime and selecting a granularity value from among the plurality of different granularity values to assign the pattern of downtime.