Evaluation Of Economic Forecasting-model By Entropy

Forecasting predicts a target for a future specific value or a trend by a quantificationally way or qualitative analysis through historical dates. Forecast has been widely used in every area of society. Forecast-model has a profoundly and extensively develop, accompanying technology advanced. Many models has been widely gradually accepted by us, like Regression Model, Synergetic Model, Grey Prediction, Phase Space Reconstruction, Artificial neural network, Markov forecast, Time series analysis. Evaluation of forecasting-model has become a key problem in this circumstance.At this time we evaluate model through this way that calculate the degree of deviation of historical date deviated from predicted value. This method of evaluate model obviously miss some points, this paper's main job is to solve these shortages.Entropy is uncertainty of random variables by mathematically. Entropy has been widely used in nature science area and social science area. In this paper we use entropy to evaluate the forecasting-model. We measure the entropy of dependent variable by measuring the entropy of transformation of independent variable. This paper first bring the problem we need to study, then make a literature survey, interpreting the method and way; following next some definition of relative concept is presented, including forecasting-model introduction, entropy interpretation, function of random variable; then analyze the useful of entropy in evaluation of forecasting-model when dependent variable follow the discrete distribution and continuous distribution; conclude the effect factor of entropy of independent variable, provide some suggestion of evaluate forecasting-model. At last this paper concludes the research and brings some advices for the further research.This paper can help comprehend of forecast and forecasing-model. It brings out a new angle to evaluate the models. If other conditions won't change, accompanying the smaller of the entropy of dependent variables, the model is better. At the,same time also provide a way of choosing the independent variable. If the independent variable can make the entropy of dependent variable smaller, then it is a better independent variable.