Purpose and area of application

STATFOR (STATistical FORecasting) package is intended for statistical forecasting (prediction) of quantitative and non-numerical variables and has the following functions:

• data structure analysis which allows to choose the most adequate model for forecasting and includes missing values processing, polynomial and median smoothing, filtering, outliers and “change-points” detection, seasonality exclusion, detection of hidden periodicities, and trend presence testing;
• short-term forecasting of time series with exponential smoothing methods;
• forecasting of non-stationary time series using a trend model from the given family of parametric models;
• forecasting of arma stationary time series;
• robust regression forecasting according to multiple linear regression model;
• non-parametric regression forecasting;
• forecasting by logistic regression model.
• forecasting by linear and quadratic robust discriminant analysis models.

Technical opportunities

Certain algorithms, implemented in the package, represent the results of original scientific research of the Mathematical modeling and data analysis department of the BSU, such as Bayesian approach in forecasting according to trend and autoregression models, local median method of forecasting of time series with trend, regularized method of parameters estimation for regression models, and forecasting methods that are robust to data distortions.

The fields of applications of the STATFOR:

• prediction and recognition of signals in signal processing;
• analysis and prediction of medical and ecological data;
• forecasting of economic and financial time series, etc.

STATFOR package provides with convenient user interface as well as with built-in data editor and extended graphic functions.
Import of data in EXCEL and DBF formats is also supported.

Hardware & software requirements
• Microsoft Windows 95/98/NT 4.0 or higher
• A Pentium® or faster Intel® Processor
• HDD up to 5 mb of free space