System and method for capacity prediction
A method, computer program product, and computer system for identifying, by a computing device, historical data. Usage data may be forecasted for at least two frequencies of a plurality of frequencies associated with the historical data based upon, at least in part, the historical data. It may be determined that the usage data is forecasted more accurately for a first frequency of the at least two frequencies than for a second frequency of the at least two frequencies. The usage data forecasted for the first frequency of the at least two frequencies may be selected based upon, at least in part, determining that the usage data is forecasted more accurately for the first frequency. Future usage of an item may be predicted based upon, at least in part, the usage data forecasted for the first frequency.
1. A computer-implemented method comprising:
identifying, by a computing device, historical data;
processing the historical data, wherein processing the historical data includes one or more of identifying a null value in the historical data and removing the null value from the historical data and checking the historical data for stationarity;
forecasting usage data for at least two frequencies of a plurality of frequencies associated with the historical data based upon, at least in part, the historical data, wherein the plurality of frequencies includes a daily frequency of usage data, a weekly frequency of usage data, and a monthly frequency of usage data;
determining that the usage data is forecasted more accurately for a first frequency of the at least two frequencies than for a second frequency of the at least two frequencies, wherein determining that the usage data is forecasted more accurately for the first frequency includes running a model on the historical data for at least the daily frequency of usage data, the weekly frequency of usage data and the monthly frequency of usage data, wherein the model forecasts the usage data that occurs during the at least two frequencies;
selecting the usage data forecasted for the first frequency of the at least two frequencies based upon, at least in part, determining that the usage data is forecasted more accurately for the first frequency;
predicting future usage of an item based upon, at least in part, the usage data forecasted for the first frequency; and
providing, by the computing device, a notification of the future usage of the item.
2. The computer-implemented method of claim 1 wherein the model includes a Seasonality Trend using Loess (STL) model.
3. The computer-implemented method of claim 1 wherein determining that the usage data is forecasted more accurately for the first frequency includes quantifying performance of the model using a precision matrix.
4. The computer-implemented method of claim 3 wherein determining that the usage data is forecasted more accurately for the first frequency includes providing an upper bound and a lower bound with a mean forecast for at least the first frequency and the second frequency.
5. The computer-implemented method of claim 3 wherein the precision matrix includes a Mean Absolute Percentage Error (MAPE).
6. The computer-implemented method of claim 1 , wherein processing the data further comprises including test data within the historical data.
7. A computer program product residing on a computer readable storage medium having a plurality of instructions stored thereon which, when executed across one or more processors, causes at least a portion of the one or more processors to perform operations comprising:
identifying, by a computing device, historical data;
processing the historical data, wherein processing the historical data includes one or more of identifying a null value in the historical data and removing the null value from the historical data and checking the historical data for stationarity;
forecasting usage data for at least two frequencies of a plurality of frequencies associated with the historical data based upon, at least in part, the historical data, wherein the plurality of frequencies includes a daily frequency of usage data, a weekly frequency of usage data, and a monthly frequency of usage data;
determining that the usage data is forecasted more accurately for a first frequency of the at least two frequencies than for a second frequency of the at least two frequencies, wherein determining that the usage data is forecasted more accurately for the first frequency includes running a model on the historical data for at least the daily frequency of usage data, the weekly frequency of usage data and the monthly frequency of usage data, wherein the model forecasts the usage data that occurs during the at least two frequencies;
selecting the usage data forecasted for the first frequency of the at least two frequencies based upon, at least in part, determining that the usage data is forecasted more accurately for the first frequency;
predicting future usage of an item based upon, at least in part, the usage data forecasted for the first frequency; and
providing, by the computing device, a notification of the future usage of the item.
8. The computer program product of claim 7 wherein the model includes a Seasonality Trend using Loess (STL) model.
9. The computer program product of claim 7 wherein determining that the usage data is forecasted more accurately for the first frequency includes quantifying performance of the model using a precision matrix.
10. The computer program product of claim 9 wherein determining that the usage data is forecasted more accurately for the first frequency includes providing an upper bound and a lower bound with a mean forecast for at least the first frequency and the second frequency.
11. The computer program product of claim 9 wherein the precision matrix includes a Mean Absolute Percentage Error (MAPE).
12. A computing system including one or more processors and one or more memories configured to perform operations comprising:
identifying, by a computing device, historical data;
processing the historical data, wherein processing the historical data includes one or more of identifying a null value in the historical data and removing the null value from the historical data and checking the historical data for stationarity;
forecasting usage data for at least two frequencies of a plurality of frequencies associated with the historical data based upon, at least in part, the historical data, wherein the plurality of frequencies includes a daily frequency of usage data, a weekly frequency of usage data, and a monthly frequency of usage data;
determining that the usage data is forecasted more accurately for a first frequency of the at least two frequencies than for a second frequency of the at least two frequencies, wherein determining that the usage data is forecasted more accurately for the first frequency includes running a model on the historical data for at least the daily frequency of usage data, the weekly frequency of usage data and the monthly frequency of usage data, wherein the model forecasts the usage data that occurs during the at least two frequencies;
selecting the usage data forecasted for the first frequency of the at least two frequencies based upon, at least in part, determining that the usage data is forecasted more accurately for the first frequency;
predicting future usage of an item based upon, at least in part, the usage data forecasted for the first frequency; and
providing, by the computing device, a notification of the future usage of the item.
13. The computing system of claim 12 wherein the model includes a Seasonality Trend using Loess (STL) model.
14. The computing system of claim 12 wherein determining that the usage data is forecasted more accurately for the first frequency includes quantifying performance of the model using a precision matrix.
15. The computing system of claim 14 wherein determining that the usage data is forecasted more accurately for the first frequency includes providing an upper bound and a lower bound with a mean forecast for at least the first frequency and the second frequency.