Statistics and forecasts

With QMF, you can run a variety of statistical analyses against query results and use the returned information to identify trends and predict future events.

QMF for Workstation forecasts

With forecasting capabilities, you can make projections of future values based on past values. Using forecasts, organizations can prepare for changes in economic or competitive conditions by analyzing time series historical data to predict performance and future trends. For example, in a supply chain, if the forecast demand matches the actual demand then significant efficiencies can be achieved in terms of production, distribution, and return.

Using QMF for Workstation forecasts you can apply various predictive methods based on mathematical algorithms that model the future demand based on time series historical data that can be sourced from queries and tables containing date or time columns. The overall objective is to choose a time series method that produces a best fit model of past values, by identifying existing patterns in the data and projecting the model into the future to generate the forecast.

The following methods can be used to forecast future values:
  • If the time series is relatively stationary with no overall tendency to fluctuate at one part of the series as compared to another part of the series, then Moving Average, Weighted Moving Average, or Single Exponential Smoothing provide the best fit model.
  • If the time series has a trend with a consistent upward or downward movement over time, then Double Exponential Smoothing provide the best fit model.
  • If the series has a trend and seasonality with a pattern of peaks and troughs that repeat themselves over a time-frame of usually less than or equal to a year, then Holt-Winters method provide the best fit model.
  • If the series has a trend, seasonality and cyclicity with a pattern of peaks and troughs that repeat themselves over an extended time-frame usually greater than a year, then the Multiplicative Decomposition method provide the best fit model.
  • If the series displays none of the above, then Neural Networks be used to mathematically fit the historical data.
  • If there are theoretical reasons to indicate that the data should follow a clear mathematical function, then one of the Curve fitting methods can be used.

In addition to the above methods, the forecaster is also able to manually adjust any predicted values based on the forecaster's knowledge and any external events.

As most new users discover, the ability to quickly plot and compare each forecast method is a major feature of QMF forecasts. However, a forecaster's knowledge and experience help to reduce the possibilities and consequently provide greater confidence and reliability in the forecast.

Statistics and forecasts

QMF Analytics for TSO includes statistical analysis and forecasting capabilities that you can run against query results in the TSO environment. You can take the output of a statistical analysis or forecast, such as a graph, a chart, or a table of data, and use it as a means to visualize, validate, and understand the patterns behind your data.

The formulas and calculations associated with statistical analysis and forecasting can be complex. As a business user, you might think that the benefits provided by statistical analysis and forecasting are outweighed by the time and effort required to make successful use of them. But QMF Analytics for TSO provides quick-to-use statistics for business professionals, as well as powerful statistical analysis and forecasting capabilities for the experienced statisticians.

QMF Analytics for TSO applies forecasting techniques such as Box-Jenkins modeling, exponential smoothing, moving averages, regression analysis, and trend projection.