A Neural Network Model for Time-Series Forecasting

The paper presents some aspects  regarding the use of pattern recognition  techniques and neural networks for the activity  evolution diagnostication and prediction  by means of a set of indicators. Starting from the  indicators set there is defined a measure  on the patterns set, measure representing a  scalar value that characterizes the  activity analyzed at each time moment. A pattern  is defined by the values of the  indicators set at a given time. Over the classes set  obtained by means of the classification  and recognition techniques is defined a  relation that allows the representation  of the evolution from negative evolution towards  positive evolution. For the  diagnostication and prediction the following tools are  used: pattern recognition and multilayer perceptron. The paper also presents the  REFORME software written by the authors and the results of the experiment  obtained with this software for  macroeconomic diagnostication and prediction during the  years 2003-2010.