Forecasting Performance - HPE XP7 User Manual

Forecasting performance

Guidelines for selecting data range to receive an optimal forecast
To validate the forecasted data, we need to understand the trend of the existing data, as the forecasted
data is an extension of the existing trend. For example, in the following figure, the forecasted data
represents a trend of the ThP pool occupancy values and not the actual values. The following graph
indicates the trend of the actual values. The forecasted values is an extension of the trend of the selected
data points.
The following graph indicates the forecasted graph in red. The forecasted data starts below the actual
value and gradually grows upwards, based on the trend of the actual data.
The accuracy of the forecasting depends on the following factors:
• Data range size: The accuracy of the forecast is directly proportional to the size of data chosen. So,
select a range as large as possible. HPE recommends to have at least 1 day of performance data for
Daily forecast, at least 1 week of performance data for weekly Forecast and at least 1 month of
performance data for monthly forecast for PA to project an optimal forecast.
• End date of the selection: The end date must be near to the current date. If the end date is not close to
the current date, you will get a forecasted value for the performance data that is already available.
• No variance: Select a data range that has at least some variance. If the selected data range has
constant values for most of the range, the forecast may follow the constant data pattern.
• Empty collection ranges: Missing data points may induce error in the forecasted data.
Forecasting performance 353