The Improvements Of GMDH Algorithm And Research Of The Prediction And Early Warning On Coal Market System

The coal energy is an important means of production which related to China’s national economyand people’s livelihood, and plays an active and important role in the national economic construction.In recent years, the serious imbalance on the proportion of supply and demand of the coal market, andfrequent price fluctuations, especially soaring or plummeted coal prices, severely handicaped thehealthy development of the coal industry, had had a negative impact on national economicdevelopment, great damages on social stability and people’s lives. Therefore, in order to protect thethe coal market to keep normal operation of the coal market, doing research about forecast and earlywarning of coal market system has a very important practical significance.In this paper, trying to use GMDH method to do some research of predictiction and the warningof the coal market system regarded as the research object of this paper. First, the paper presentsseveral improvements of GMDH algorithm: the proposed "symbolic expression method" and the"weight coefficient vector method" are used to obtain the explicit output expr ession of GMDHnetwork directly; the input variables (i. e., the first lay er nodes) are always included in the candidatepool of the partial polynomials for determining the construcution of the higher layers (which is namedas "adding initial variables method"), which can effectively prevent the loss of useful information;based on the idea of integrated learning, presenting a kind of selective integration algorithm of agroup of candidate GMDH network individuals (which were different each other) developed withpunitive division of the samples, whose optimal subset of the candidate GMDH models was obtainedby using the genetic algorithm; providing a gadget named as "GMDH modeler with explicitexpressions" by using GUI programming of Matlab. Then, the paper designs the block diagram ofstructure of the warning of the coal market system, builds three forecast models of coal prices, supplyand demand by using the above gadget. Obtains the predicted values of each factor included in theabove three prediction models by using BP neural network and GMDH method, and further obtainsthe predictive values of coal price, supply and demand for2008year. Finally, the paper indicates thatChina’s coal market in2008is in a “heavy alarm” state and is sended out orange warning light bycombining predicted values of coal price fluctuations and the proportion of the coal supply anddemand for2008year with the warning proposal of coal market system. The results of this predictionand early warning are basically consistent with the actual running status of coal market system in2008, which shows the GMDH method is an effective method for the early warning of the coal marketsystem.