Most historical or time-based data contains an underlying trend or seasonal pattern. However, most historical data also contains random fluctuations (“noise”) that make it difficult to detect these trends and patterns without a computer. Predictor uses sophisticated time-series methods to analyze the underlying structure of the data. It then projects the trends and patterns to predict future values.
Predictor uses two types of forecasting:
Time-series forecasting breaks historical data into components: level, trend, seasonality, and error. Predictor analyzes these components and then projects them into the future to predict likely results.
Multiple linear regression works best when outside influences have an effect on the variable that you want to forecast. Regression takes historical data from the influencing variables and determines the mathematical relationship between these variables and the target variable. It then uses time-series forecasting methods to forecast the influencing variables and combines the results mathematically to forecast the target variable.
In Predictor, a data series is a set of historical data for a single variable. When you run Predictor, it uses each time-series method on each of the selected data series and calculates a mathematical measure of goodness-of-fit. Predictor selects the method with the best goodness-of-fit as the method that will yield the most accurate forecast. Predictor performs this selection automatically, but you can also select individual methods manually or override the method that Predictor recommends with a different one.
The final forecast shows the most likely continuation of the data. Keep in mind that all these methods assume that some aspects of the historical trend or pattern will continue into the future. However, the farther out you forecast, the greater the likelihood that events will diverge from past behavior, and the less confident you can be of the results. To help you gauge the reliability of the forecast, Predictor provides a confidence interval indicating the degree of uncertainty regarding the forecast.
After finding the best forecast for the data, Predictor displays detailed output that can include statistics, charts, reports, and interactive Microsoft Excel PivotTables. Predictor can also paste the forecasted values into a spreadsheet and create Crystal Ball assumptions from forecasted values so you can perform a “what-if” simulation.
The following topics describe how to set up Predictor forecasts using default settings so you can generate results quickly for further analysis:
Predictor basics are demonstrated in Tutorial 1—Shampoo Sales. You may find it helpful to work through this tutorial now, or read through the following sections first and then try the tutorial. When you are ready to expand your forecasting skills, Setting Up Predictor Forecasts, provides detailed instructions.